Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: ⁠⁠⁠⁠⁠⁠⁠https://substack.com/@thespecialistgeneralist⁠⁠⁠⁠⁠⁠⁠. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #9: Delve into the pressing question of AI data center viability in the UK, exploring how Special Compute Zones and nuclear power reforms could position the nation as a global leader in AI infrastructure. Luis von Ahn of Duolingo shares his take on gamification versus learning efficacy and how AI-powered tools are reshaping education. Gwern Branwen offers thought-provoking insights on anonymity, AI scaling, and the future of AGI. Examine the growing debate around AI scaling laws, the progress (and limits) of large language models, and innovative use cases like AI translating whale communication patterns. Startups like Odyssey are redefining visual storytelling through Hollywood-grade generative AI tools, while breakthroughs in genetic design, venom-based medicine, and nuclear fusion push the boundaries of science. Also read more for China’s shifting influence in Hong Kong’s financial sector and Goldman Sachs’ power struggles over its China partnership. Finally, look at cultural shifts in communication, the history of non-smiling portraits, and groundbreaking discoveries like the frozen body of a 35,000-year-old saber-toothed kitten.

Quotations in italics

Table of contents:

1. Top things I’ve read this week

2. AI/Technology news of the week

3. Interesting AI use cases of the week

4. Start-Up news of the week

5. Science news of the week

6. World news of the week

7. USA news of the week

8. China news of the week

9. UK news of the week

10. Academic paper(s) of the week

11. Culture of the week

12. Miscellaneous

1. Top things I've read this week

Getting AI datacentres in the UK - Inference Magazine

If trends in computational power scale-up continue, by 2030, training the largest model would use 2.15 times the UK’s entire energy generation capacity.

As things stand, no developer would choose to build an AI datacentre in the UK. The UK’s nuclear construction costs are 4.5 times higher than South Korea, and building reactors takes twice as long. Approving a new power plant has taken 6 and 12 years in the last two instances, while approving a reactor with the same basic design in France took just 3 years. If a datacentre operator wanted to connect to the grid instead of building new power, it would take up to 15 years to get a grid connection, and the industrial electricity prices are four times higher than in the US and 45% higher than in France.

Despite these facts, the UK can make a series of reforms and, in just a few weeks of swift action, the UK can become the best place in the world to build AI datacentres and the nuclear power to support them. Special Compute Zones would create an alternative planning and regulatory approval process for nuclear power, AI datacentres, and the transmission and networking infrastructure to support them. If these reforms allowed UK projects to close the cost gap with South Korea by two thirds, nuclear power would become 37% cheaper than an equivalent blend of renewable power for datacentres. With high levels of speed and certainty during the approval process, the UK has the opportunity to have catch the wave of AI infrastructure development.

• This a really important essay on how by creating Special Compute Zones and reforming nuclear energy policies, the UK can position itself as a global leader in AI infrastructure. This initiative would not only meet the burgeoning energy demands of AI development but also drive economic growth, enhance national security, and ensure long-term competitiveness in the rapidly evolving field of artificial intelligence.

Duolingo CEO Luis von Ahn on AI, gamification, and the power of freemium - The Verge

The cofounder of the world’s largest education app thinks AI and gamification can supercharge language learning.

• In the interview, Luis von Ahn, CEO of Duolingo, discusses the integration of AI to enhance language learning through features like AI-powered conversations and interactive adventures, emphasizing that engagement is crucial for effective education. He acknowledges the tension between gamification and learning efficiency but argues that keeping users motivated is essential since "you can't teach somebody who’s not there." Luis also explains that while the majority of Duolingo's users are on Android devices, most of the revenue comes from iOS users in wealthier countries, effectively subsidizing free education for users in developing nations. This highlights the company's challenge in balancing its mission to provide accessible education with the need for profitability. Additionally, he touches on organizational strategies, such as a metrics-driven approach and in-person collaboration, to maintain high-quality content and innovation within the company.

• I was particularly interested by his commentary on this topic of gamification and engagement and how that interacts with the fundamental learning mission:

  • The hardest thing about me trying to teach you something is just keeping you engaged. Somehow, people forget. I see some people saying, “You can learn quantum physics with ChatGPT.” And yeah, sure, but that’s just not that impressive. You can learn quantum physics with a book. The technology to learn that has been around for a long time. It’s called a book and it works. It’s just that people don’t really want to read a quantum physics book. And similarly, most people probably don’t want to go to ChatGPT and start asking questions about quantum physics. It’s the same thing for language learning. Large language models are very good at getting you to practice, but keeping you engaged is pretty hard. I don’t know if large language models are going to help all that much with that part.

  • In the end, this is a sad thing, but the reality is that Duolingo is very gamified. I wholeheartedly believe most people would rather spend more time playing Candy Crush than talking to others. That’s maybe a sad truth. And there are some exceptions. I mean, people love talking to someone they’re in love with, and sure, that’s nice, but the reality is, most of the time, most people would rather spend time playing Candy Crush or scrolling on Instagram than talking to others. I just don’t think large language models are going to help much with all of that.

  • From the beginning, this is a central thesis that we believe here at Duolingo: the hardest thing about learning something by yourself is staying motivated. In fact, that is probably the reason for the vast majority of our success is that we realized that early on. From the beginning, we have tried to have a thing that is enjoyable to use and that keeps you coming back. We have probably spent more effort on that than anything else. Internally, our feeling is that learning a language is a lot like working out. It doesn’t matter all that much whether you’re doing the elliptical or a Peloton or a treadmill. By far, what matters the most is that you’re doing it every day, whatever the hell you’re doing. It’s kind of the same with Duolingo. Maybe some methods are more efficient than others, but what matters is that you’re doing it every day. We got very good at that. Now, once we got very good at that, we started trying to add more sophistication in what we teach, and we’ve been doing that for the last few years. But always, primarily, we are a motivation engine.

  • Do you think that there’s a conflict between gamification and engagement — the things that you’re historically successful at — and education?

    • Yes, there is.

  • How do you manage that conflict?

    • Very easily. Always go with engagement.

    • I’ll give you many arguments, but the one that works the most is this: It doesn’t matter how effective you are. You can’t teach somebody who’s not there. That’s it. People leave. The reality is it’s not always true that engagement and learning outcomes are at odds. But when they are, we usually prefer going for engagement.

Gwern Branwen - How an Anonymous Researcher Predicted AI's Trajectory - Dwarkesh Patel Podcast

• In this conversation, Gwern Branwen, an anonymous researcher and writer known for his deep dives into various subjects, discusses the multifaceted benefits of anonymity, highlighting how it prevents preconceived biases and allows content to be evaluated on its own merits. He explores the future of AI and automation, suggesting that while automation may start from the bottom up, human oversight, particularly in vision and long-term strategy, will remain crucial at the top levels of organizations. Gwern reflects on his early recognition of AI scaling laws and emphasizes the significance of compute power and data over algorithms in AI development. He shares his thoughts on AGI timelines, considering the rapid advancements in AI and contemplating how to best utilize the remaining years before AGI's potential arrival—focusing on human elements that AI cannot replicate, such as personal insights and experiences. Gwern delves into his personal penchant for "rabbit holes," or intensive explorations of topics, which have been a driving force in his work. Gwern also touches on his time as a Wikipedia editor, his transition to personal blogging, and his unique work habits, including how online discussions fuel his motivation to write. Finally, Gwern expresses his hopes for answers to significant questions by 2050 and discusses the potential diversity of AI minds compared to human intelligence.

2. AI/Technology news of the week

AI and The Last Mile

Discerning human judgement as a luxury good

So here’s an uncomfortable truth: the "last mile" will become the new gated community, where inside the velvet ropes are expert humans with discernment. While AI promises to democratize everything from college admissions to job hunting, this human discernment will become a luxury good. The wealthy parent hires the counselor who "just knows" which college is the right fit. The well-connected job seeker gets the coffee chat that no algorithm can replicate. Even as AI gets better at matching people with opportunities, those crucial final judgment calls — the ones that often make all the difference — will remain hidden behind the velvet rope of privilege and social capital. We're creating a world where AI algorithms serve the majority while human insight becomes the ultimate premium service.

• In the essay, the author argues that as AI increasingly handles decision-making tasks, the nuanced human judgment required for the "last mile" of complex decisions becomes a luxury accessible only to those who can afford it. This "last mile" refers to the final 5-15% of context-dependent understanding that algorithms cannot provide—such as interpersonal dynamics in hiring or the subtle "feel" of a neighborhood when buying a house. The author suggests that while AI democratizes access to data-driven services, it simultaneously elevates human discernment into a premium commodity, creating a new form of inequality. Instead of eliminating middlemen, AI shifts the value to human experts who can offer the nuanced insights that algorithms lack, thereby making discerning human judgment the ultimate premium upgrade in an AI-dominated world.

OpenAI Nears Launch of AI Agent Tool to Automate Tasks for Users

OpenAI is preparing to launch a new artificial intelligence agent codenamed “Operator” that can use a computer to take actions on a person’s behalf, such as writing code or booking travel

• Anthropic unveiled a similar agent that can process what’s happening on the user’s computer in real time and take actions on their behalf. OpenAI-backer Microsoft Corp. also recently launched a set of agent tools designed to send emails and manage records for workers. And Alphabet Inc.’s Google is said to be preparing to release an AI agent, according to The Information.

Some small signs that AI progress is slowing down - from Noahopinion’s At least five interesting things: Trump aftermath edition (#52)

In recent weeks, a number of people have started to claim that the benefits of scaling are slowing down. The Information reported that OpenAI’s new models show more incremental improvements than their previous ones. Reuters reported that Ilya Sutskever, one of the world’s leading LLM researchers, is pessimistic that scaling will be enough to sustain rapid progress

• Noah discusses the idea that large language models (LLMs) might no longer yield substantial progress, challenging the widely-held "scaling hypothesis" that AI will inevitably advance by increasing model size, data, and compute power. While scaling has indeed driven AI breakthroughs in the past decade, recent evidence suggests diminishing returns. OpenAI's latest models show incremental rather than revolutionary improvements, and high-quality data sources are drying up, limiting future scaling potential. Additionally, pervasive issues like "hallucinations" — where models confidently generate false information — persist despite scaling efforts. This indicates that LLMs may be asymptotically biased, inherently unable to reliably discern truth from falsehood, which raises doubts about "scaling to AGI." Consequently, the path to achieving AGI may require novel approaches beyond simply expanding existing models, making the future of AI development more complex and uncertain than optimists might expect.

• Platformer has more on this here:

• So what’s happening?

  • One is that AI companies are running out of high-quality new data sets to feed into their models…Only recently did OpenAI, Meta and others begin paying to license high-quality data from publishers and other sources. Efforts to paper over this deficiency with synthetic data — data made by the AI labs themselves — have yet to generate a breakthrough.

  • Another, potentially more troublesome explanation is that superintelligence simply can’t be built with LLMs alone. Progress in AI is measured primarily in the ability of these models to hit various benchmarks. And while LLMs have been crushing benchmarks at impressive speed, the benchmarks can be gamed. More importantly, models fail to understand cause and effect and other tasks that require multi-step reasoning. They often make factual mistakes and lack real-world knowledge.

  • Perhaps the issue, as Tyler Cowen writes at Marginal Revolution today, is that we have taken too blinkered a view of “knowledge” — and placed impossible expectations on LLMs. “Knowledge is about how different parts of a system fit together, rather than being a homegenous metric easily expressed on a linear scale,” he writes. “ There is no legible way to assess the smarts of any single unit in the system, taken on its own. Furthermore, there are many ‘walls,’ meaning knowledge is bumpy and lumpy under the best of circumstances. It thus makes little sense to assert that an entity is ‘3x smarter’ than before.”

• This all being said

  • Sam Altman posted “There is no wall,” on X.

  • The Information reported that the company’s recent focus on its reasoning-focused o1 model represented one way the company is attempting to accelerate its progress again.

    • According to Noam Brown, a researcher at OpenAI who worked on o1, it uses a technique known as “test time compute,” which effectively gives the model more time and compute at the moment of inference. “It turned out that having a bot think for just 20 seconds in a hand of poker got the same boosting performance as scaling up the model by 100,000x and training it for 100,000 times longer,” he said at TED AI.

  • Anthropic CEO Dario Amodei appeared on Lex Fridman’s podcast this week, and offered a similarly upbeat take.

    • “We have nothing but inductive inference to tell us that the next two years are going to be like the last 10 years,” said Amodei, who co-authored that first paper on scaling laws while at OpenAI. “But I’ve seen the movie enough times, I’ve seen the story happen for enough times to really believe that probably the scaling is going to continue, and that there’s some magic to it that we haven’t really explained on a theoretical basis yet.”

The Bitter Religion: AI’s Holy War Over Scaling Laws - The Generalist

• On the same topic as above here is useful overview of the ideological divide over scaling.

• Central to this debate is Richard Sutton's "The Bitter Lesson," which advocates for the idea that scaling computational power—rather than refining specialized techniques—is the key driver of AI advancement. Proponents of this view, like OpenAI's Sam Altman, assert that exponentially increasing computation will inevitably lead to AGI, reinforcing a doctrine that has become almost religious in its adherence. Opponents, including figures like Ilya Sutskever, argue that this approach is reaching its limits and that novel scientific research is necessary for further breakthroughs. This ideological split has significant implications for the future of AI, affecting investment strategies, research directions, and the broader technological landscape. By drawing parallels between religious fervor and the convictions held by both camps, the article questions whether the current faith in scaling laws is sustainable or if a paradigm shift is imminent in the quest for true artificial intelligence.

Austrian economics and AI scaling - Marginal Revolution

Now in the last two days we have seen various media accounts (Bloomberg), some citing AI experts, claiming that scaling has slowed down or is no longer working. Put aside whether these claims are correct. I remain bullish on AI progress, because the AI world is showing so many signs of significant ferment. Ask yourself for instance — is it attracting the best minds and most ambitious people? Just as the postulation of AI scaling never so much excited me, neither does talk of its possible diminution much discourage me.

It is still a question of whether and at what pace we can find or create “AI-shaped holes” in organizations, or with individuals. And that is up to us.

• Tyler reflects on his skepticism toward predictions about AI progress based solely on "scaling laws," drawing from thinkers like Hayek and Polanyi who argue that knowledge is complex, non-linear, and cannot be measured on a simple scale. He suggests that both AI pessimists and optimists misunderstand the nature of knowledge by assuming it can be quantified in such a way. Instead, he emphasizes the importance of social and historical contexts that have fostered past artistic and scientific revolutions—environments rich in talent, competition, and innovation. Observing similar conditions in the current AI landscape, with significant ferment and attraction of ambitious minds, he remains optimistic about AI's future. Ultimately, he believes that the pace and impact of AI progress depend on our ability to create "AI-shaped holes" in organizations and society, integrating the technology in meaningful ways.

Ep17. Welcome Jensen Huang | BG2 w/ Bill Gurley & Brad Gerstner

• This conversation is an in-depth interview with Jensen Huang, the CEO of NVIDIA, conducted by investors Brad and Clark from Altimeter Capital. The discussion revolves around the rapid advancements in artificial intelligence (AI), NVIDIA's role in this transformation, the company's competitive advantages, and the future implications of AI on various industries and society at large.

• Here are some key takeaways:

  • Timeline to a personal assistant in our pockets—an Artificial General Intelligence (AGI) that knows everything about us and can perform complex tasks like booking appointments. Jensen believes this is imminent in some form and will improve over time.

  • Rate of Change: Jensen acknowledges that the current rate of change in AI is the fastest he's seen in his career. He attributes this to the reinvention of computing, where the marginal cost of computing has decreased by a factor of 100,000 over ten years, surpassing Moore's Law significantly.

  • Scaling Intelligence: The conversation touches on the importance of scaling both pre-training and inference in AI models. Jensen notes that while training large models was once the primary focus, inference—especially with iterative reasoning and reflection—is becoming increasingly computationally intensive.He also anticipates the need for millions of GPUs to handle future AI workloads. The company is preparing for this demand by innovating in system architecture and software to enable distributed and federated learning.

  • Collaboration with AWS and Others: While companies like Amazon are developing custom chips (e.g., Trainium and Inferentia), NVIDIA remains a key partner. Jensen notes that NVIDIA's platforms complement these custom solutions and are integrated into various cloud ecosystems.

  • Modernizing Data Centers: Jensen argues that existing data centers, primarily CPU-based, need modernization to handle AI workloads. This represents a significant market opportunity as companies will prefer investing in AI-optimized infrastructure.

  • Sustained Demand: Contrary to critics who suggest the current demand is a temporary boom, Jensen believes the need for AI computing infrastructure will continue to grow exponentially due to the reinvention of computing and the emergence of AI factories.

Exclusive-Amazon developing driver eyeglasses to shave seconds off deliveries, sources say

⁠Amazon is developing smart eyeglasses for its delivery drivers to guide them to, around and within buildings, as it tries to smooth the final stretch of an order's journey to a customer's home, five people familiar with the matter said. f successful, the glasses would provide drivers with turn-by-turn navigation on a small embedded screen, along their routes and at each stop, according to the people, who spoke to Reuters on condition of anonymity because the project is not public.

Such directions could shave valuable seconds off each delivery by providing left or right directions off elevators and around obstacles such as gates or aggressive dogs. With millions of packages delivered daily, seconds add up. The glasses would also free drivers from using handheld Global Positioning System devices, allowing them to carry more packages.

The project underscores the online seller's efforts to reduce delivery costs per package and support margins as it fights increased competition from Walmart, which has stepped up its e-commerce efforts and cut prices. Walmart is paying independent delivery drivers new incentives to deliver online ord⁠⁠ers during the holiday season, the retailer told Reuters.

Adding payments to your LLM agentic workflows - Stripe

With the Stripe agent toolkit you can now easily integrate Stripe into the most popular agent frameworks. This enables you to automate common workflows that depend on Stripe and also helps unlock new use-cases by providing agents access to financial services on tools. In addition, usage-based billing can quickly integrate in these agent frameworks to bill your customers.

Amazon unveils AI chip to rival Nvidia

Amazon is set to launch its "Trainium 2" chip next month as part of its strategy to reduce reliance on Nvidia and strengthen its position in the AI hardware market. The chip is developed by Annapurna Labs, which Amazon acquired in 2015 for $350 million to support its in-house chip efforts. Trainium 2 is currently being tested by Anthropic, which has received $4 billion in Amazon backing, along with Databricks, Deutsche Telekom, Ricoh, and Stockmark.

Baidu unveils Xiaodu AI glasses: World’s first AI glasses with native Chinese model support

The glasses support six core functionalities: real-time Q&A through walk-and-ask queries, calorie recognition, object identification through an encyclopedia feature, audio-visual translation, and intelligent reminders. They are also equipped with AI anti-shake technology for photos and videos, enabling users to document their experiences seamlessly.

The AI agents stack

Although we see a lot of agent stack and agent market maps, we tend to disagree with their categorizations, and find they rarely reflect what we observe actually being used by developers. The agent software ecosystem has developed significantly in the past few months with progress in memory, tool usage, secure execution, and deployment, so we decided it was time to share our own “agent stack” based on our own learnings from working on open source AI for over a year and AI research for 7+ years.

• Clink the link for more details on each of the categories.

3. Interesting AI use cases of the week

Learning to speak to whales using AI, with David Gruber (Ep. 137)

• In this episode of "Big Brains," host Paul Rand interviews David Gruber, a biology and environmental science professor at the City University of New York and founder of Project CETI (Cetacean Translation Initiative). Gruber discusses his team's groundbreaking work using artificial intelligence to decode the complex communication patterns of sperm whales. Drawing parallels to potential communication with extraterrestrial life, he explains how AI technologies can help bridge the gap between human and non-human languages. The conversation delves into the intricacies of sperm whale "codas," which are patterns of clicks that function similarly to a phonetic alphabet, indicating a sophisticated form of communication. Gruber also highlights the ethical considerations of studying marine life, emphasizing the development of non-invasive research methods and technologies inspired by nature to minimize harm to the animals. The episode underscores the potential of AI and bioacoustics to foster deeper understanding and connection between humans and other intelligent species on Earth.

The Beatles’ final song, "Now and Then," became the first AI-assisted track to receive Grammy nominations.

4. Start-Up news of the week

Odyssey

We're pioneering Hollywood-grade generative worlds, where beautiful scenery, characters, lighting, and motion can be both generated and directed. This technology will deliver the game-changing efficiencies and creativity of AI, but importantly, give you full, fine-tuned control over every element in your scene. From stunning mountains, to lush vegetation, to beautiful rivers, to epic characters, all the way to the low-level materials, lighting, motion, and more. Simply put, it's a new way to create films and games.

This capability is made possible by going layers deeper than traditional text-to-visual models. Instead of training one model that restricts you to a single input and a single, non-editable output, we're training four powerful generative models that enable fine-tuned control over each major layer of visual storytelling. Specifically, models capable of generating high-quality geometry, photorealistic materials, stunning lighting, and controllable motion. Combined, these models enable you to rapidly generate scenes and shots exactly as you dreamt them. What’s more, we’re building workflows around these models tailored for power users, with first-class integrations into existing production workflows in use in Hollywood, gaming, and beyond.

The Influence of Bell Labs - Construction Physics

• The article explores how Bell Labs' remarkable success as an innovation hub—most notably through the invention of the transistor—led other corporations in the mid-20th century to emulate its model of combining academic freedom with industrial research. Companies like IBM, Texas Instruments, Xerox PARC, Ford, and Exxon established their own research labs, granting scientists the liberty to pursue basic research without immediate concerns for profitability, aiming to replicate Bell Labs' groundbreaking achievements. However, over time, these labs shifted towards more product-focused research due to a lack of transformative breakthroughs comparable to the transistor and growing corporate reluctance to fund speculative, long-term projects without clear returns. The unique circumstances that allowed Bell Labs to thrive—such as its monopoly status and the historical context of post-WWII technological optimism—were not reproducible, leading to the decline of such unrestricted industrial research labs and suggesting that their golden era may have been an unsustainable bubble driven by specific economic and historical factors.

Europe’s Unfair Advantages - EU.VC

• Cost Advantages: One of Europe’s most tangible benefits is the lower cost of talent. While Silicon Valley wages continue to skyrocket, Europe offers access to highly skilled engineers, scientists, and developers at a fraction of the cost.

• Rich Academic Institutions: Europe is home to some of the best academic institutions in the world, particularly in fields such as AI, biotech, and other hard sciences.

• Government Support: Europe’s various governments have taken an active role in fostering innovation. While government support can be a double-edged sword - sometimes leading to excessive bureaucracy or dependency - many initiatives, including grants and tax incentives, have kickstarted venture ecosystems that are now maturing.

• Founder Resilience: European founders have had to be more resourceful and resilient. Capital in Europe has historically been harder to come by, leading to lower, more reasonable valuations compared to the inflated prices often seen in the US.

• Room for Growth: Europe’s venture capital market is still relatively unsaturated, meaning there is plenty of room for growth.

How Modern Tools Allow You to Outcompete the Most Well-Connected Investor - Data Driven VC

• Atomico’s intelligence team, comprising data analysts, software engineers, and data scientists, operates independently to serve all other teams in the firm. Their primary mission is to create a unique data asset, which they treat as a "data product" that forms the foundation for all insights across Atomico. This data-first approach enables them to provide a single source of truth across core entities—companies and people—and offers insights on aspects like funding rounds, team positions, headcount, and web traffic, aiming to support data-driven decision-making across the entire VC lifecycle.

5. Science news of the week

Meet Evo, the DNA-trained AI that creates genomes from scratch - Science

Researchers describe an AI model, schooled on billions of lines of genetic sequences, that can deduce how bacterial and viral genomes operate and use that information to design new proteins and even whole microbial genomes. The model, known as Evo, could help scientists probe evolution, investigate diseases, develop new treatments, and potentially answer a host of other biomedical questions.

Venom Helped Create Weight-Loss Drugs. What Else Could It Do For Us? - NYTimes

• Researchers found that the venom of the Gila monster, a venomous lizard native to the southwestern United States and northern Mexico, contains a molecule called exendin-4, which mimics the human hormone GLP-1 responsible for regulating insulin production and appetite. This discovery led to the development of exenatide, initially approved for treating Type 2 diabetes, and eventually paved the way for groundbreaking weight-loss drugs like Ozempic and Wegovy. The article highlights the vast potential of venom-derived peptides in medicine, noting that scientists are exploring venoms from various creatures—such as spiders, scorpions, and cone snails—for treatments of conditions like stroke, cancer, autoimmune diseases, and chronic pain. Despite challenges like species extinction and difficulties in venom collection, the study underscores the importance of biodiversity in uncovering future medical breakthroughs.

Nuclear fusion start-up claims milestone with unconventional reactor - The FT

A small start-up in New Zealand claims it has created plasma, the first step towards nuclear fusion, in under two years and for less than $10mn after experimenting with an unconventional reactor design…. The start-up has been funded by New Zealand investors to date, but has a goal of raising a series A investment round in the first quarter of 2025. Mataira said the company would ultimately need between $500mn and $1bn to prove out all technical risks.

6. World news of the week

People want competence, seemingly over everything else - Strange Loop Cannon

• This piece critiques the inefficiencies and disconnects within U.S. governance, arguing that while policymakers are often focused on designing policies, the public prioritises competent execution.

  • Politicians love policy but people love execution. People prefer “competent” government over any other adjective, whether it’s “bureaucratic” or “rigid” or “big” or sometimes even “democratic”.

• The “Machine” – the bureaucratic structure – was created to serve the people but has become cumbersome, blocking progress and fueling public frustration. Citing examples like California’s high-speed rail delays, costly rural broadband projects, and regulatory overreach, the author suggests that these dysfunctions foster distrust and political disillusionment. They call for a politics focused on state capacity, pragmatic governance, and removing obstacles to impactful action, advocating a politics centered on competence over ideology.

  • Bad implementation is endemic. It’s everywhere. State capacity is anaemic. And until it can be fixed, there can be no party of competence. Only parties shouting at each other about who created which mess instead of cleaning anything up….

    It's the only way to bring some positive energy to politics. Not about right or left or ideologies writ in stone, but about competence. Building something meaningful and bulldozing what's in your way to get it done. This is a strong positive vision of what the world could be, and it needs a champion. We should embrace it.

The 35,000 MW Dream and the Indonesian Energy Trajectory

What the Jokowi era electrification efforts reveal about the factors shaping future development in Indonesia

• Joko Widodo’s presidency in Indonesia saw ambitious infrastructure efforts, particularly in the “35,000 MW dream,” aimed at expanding the country’s electricity generation to fuel economic growth. This plan encountered significant hurdles that reveal broader dynamics shaping Indonesia's development trajectory. Key factors include regulatory constraints and PLN’s state monopoly, which impacted private sector participation, while foreign, especially Chinese, financing and construction were critical to scaling projects, often favoring coal—a low-cost but high-pollution option. Financing strategies and the viability of clean energy alternatives like geothermal and hydropower are shaped by both Indonesia’s reliance on subsidies and geopolitical considerations, such as the Belt and Road Initiative and the Just Energy Transition Partnership. Additionally, electrification efforts heavily concentrate on Java, emphasising the urban core over rural or remote regions, despite Jokowi’s symbolic push for decentralisation with the new capital, Nusantara. These development patterns underscore that political incentives, regulatory environments, and the interplay of local and international financing are pivotal in charting Indonesia's future in energy and infrastructure. The next administration’s approach will determine whether Indonesia can shift toward more sustainable and inclusive growth or remain reliant on traditional, urban-focused energy policies

7. USA news of the week

How to Build the Future of AI in the United States

• This is the second piece in Compute in America: Building the Next Generation of AI Infrastructure at Home, examining the computing infrastructure buildout needed to support next-generation artificial intelligence systems.

• The series analyzes the technical challenges of constructing massive AI data centers, projects future computing needs based on AI development trends, and proposes policy solutions to ensure this critical infrastructure is built securely in the United States.

Military’s UFO-Hunting Aerial Surveillance System Detailed In Report

The Pentagon provided new details today about how its deployable, readily reconfigurable suite of sensors called GREMLIN works to help set the stage for figuring out what unidentified objects in our skies are and are not, if they appear at all.

In its annual report released on Thursday, the All-domain Anomaly Resolution Office (AARO) shared a graphic that gives us our best look yet at what its GREMLIN system is. It was developed by Georgia Tech Research Institute specifically to help gather data about so-called unidentified anomalous phenomena, or UAPs. That’s the DoD’s new parlance for what used to be called UFOs.

8. China news of the week

Inside Goldman Sachs’ years-long power struggle over its China venture - The FT

• In the early 2000s, Goldman Sachs partnered with Chinese investment banker Fang Fenglei to access China's financial markets, anticipating the arrangement would be temporary until they could own their operations outright. Instead, the partnership lasted nearly two decades, during which Fang leveraged his connections and position to gain significant economic benefits, while Goldman struggled to gain full control of its China business. Complex ownership structures and regulatory challenges allowed Fang to maintain substantial influence, ultimately forcing Goldman to pay far more than expected to unwind the deal. This saga illustrates the difficulties foreign companies face when navigating China's financial landscape and partnering with well-connected local figures.

How Beijing Took Control of Hong Kong’s Financial Hub—and Left the West Behind

• The article highlights a significant shift in Hong Kong's financial sector, where Chinese institutions have increasingly taken dominance as Western banks and corporations retreat, altering the city's traditional role as a bridge between East and West. This transformation is evident in changes like official speeches at the Hong Kong Stock Exchange being delivered in Mandarin and the reduced participation of foreign banks in initial public offerings—dropping from involvement in roughly half of IPOs two years ago to just one-fifth currently. Chinese banks have become top earners in debt-capital markets, often engaging in higher-risk lending within China, while Western financial institutions face declining revenues and layoffs. The shift reflects China's growing economic influence and has broader implications for global finance, prompting Western banks to reassess their strategies in Asia amid evolving U.S.-China relations and concerns about Hong Kong's future as an international financial hub.

9. UK news of the week

Cambridge spends £50m to attract the brightest minds in the world - The Times

Cambridge University is launching a £50 million drive to attract the brightest minds in the world as PhD students and future Nobel laureates.

Trinity and the university will each commit £2.4 million for the first five years, with matched funding from external partners sought to meet the scheme’s overall goal of raising £50 million for new PhDs in ten years.

After the initial five years’ funding, the scheme will continue for at least a further five years, fully funded by the university to a minimum value of £4.8 million annually.

Don't switch off clean power

• Britain’s once-thriving nuclear power industry faces a critical juncture as most Advanced Gas-cooled Reactors (AGRs) are set to close by 2028, risking a complete nuclear grid shutdown and increased reliance on expensive, carbon-emitting natural gas. Extending the life of AGRs, alongside bringing Hinkley Point C online, is crucial to meeting the "Clean Power by 2030" target, reducing emissions, and lowering energy bills. Despite challenges with AGR design flaws, experts suggest extensions for Heysham B and Torness are feasible, but past closures like Hinkley Point B highlight poor policy decisions. The article calls for government intervention to balance safety concerns with the broader benefits of nuclear energy for energy security, emissions reduction, and cost savings.

10. Academic Paper(s) of the week

Digital Distractions with Peer Influence: The Impact of Mobile App Usage on Academic and Labor Market Outcomes - NEBR

Abstract: Concerns over the excessive use of mobile phones, especially among youths and young adults, are growing. We present, to our knowledge, the first estimates of both behavioral spillover and contextual peer effects, as well as the first comprehensive evidence of how own and peers’ mobile app usage affects academic performance, physical health, and labor market outcomes. Our analysis leverages administrative data from a Chinese university of three cohorts of students over up to four years merged with mobile phone records, random roommate assignments, and a policy shock that affects peers’ peers. App usage is contagious: a one s.d. increase in roommates’ in-college app usage raises own app usage by 5.8% on average, with substantial heterogeneity across students. High app usage is detrimental to all outcomes we measure. A one s.d. increase in app usage reduces GPAs by 36.2% of a within-cohort-major s.d. and lowers wages by 2.3%. Roommates’ app usage exerts both direct effects (e.g., noise and disruptions) and indirect effects (via behavioral spillovers) on GPAs and wages, resulting in a total negative impact of over half the size of the own usage effect. Extending China’s minors’ game restriction of three hours per week to college students would boost their initial wages by 0.9%. Using high-frequency GPS data, we identify one underlying mechanism: high app usage crowds out time in study halls and increases late arrivals at and absences from lectures.

• Basically Xi Jinping’s 2019 restriction of video game usage for minors improved both academic performance and labor market outcomes.

How Do You Say Your Name? Difficult-to-Pronounce Names and Labor Market Outcomes

Abstract: We test for labor market discrimination based on an understudied characteristic: name fluency. Analysis of recent economics PhD job candidates indicates that name difficulty is negatively related to the probability of landing an academic or tenure-track position and research productivity of initial institutional placement. Discrimination due to name fluency is also found using experimental data from prior audit studies. Within samples of African Americans (Bertrand and Mullainathan 2004) and ethnic immigrants (Oreopoulos 2011), job applicants with less fluent names experience lower callback rates, and name complexity explains roughly between 10 and 50 percent of ethnic name penalties. The results are primarily driven by candidates with weaker résumés, suggesting that cognitive biases may contribute to the penalty of having a difficult-to-pronounce name.

AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably - Nature

Abstract: We conducted two experiments with non-expert poetry readers and found that participants performed below chance levels in identifying AI-generated poems (46.6% accuracy, χ2(1, N = 16,340) = 75.13, p < 0.0001). Notably, participants were more likely to judge AI-generated poems as human-authored than actual human-authored poems (χ2(2, N = 16,340) = 247.04, p < 0.0001). We found that AI-generated poems were rated more favorably in qualities such as rhythm and beauty, and that this contributed to their mistaken identification as human-authored. Our findings suggest that participants employed shared yet flawed heuristics to differentiate AI from human poetry: the simplicity of AI-generated poems may be easier for non-experts to understand, leading them to prefer AI-generated poetry and misinterpret the complexity of human poems as incoherence generated by AI.

11. Culture news of the week

The 6 New Rules of Communicating - Ted Gioia

Here are the six new rules of engagement—for politicians, broadcasters, and all aspiring experts, decision-makers, and leaders.

1. You gain more trust when seated, not standing.

2. Don’t speak at people—speak with them.

3. An informal tone is more persuasive now. Even leaders must adjust to this.

4. Conversations have more influence than speeches.

5. Spontaneous communications delivered from a personal standpoint are considered more ‘real’ than a script created by a team or speechwriter.

6. Soundbites and talking points are less impactful than storytelling, humor, and off-the-cuff comments.

• Ted explores the shift in communication styles, emphasizing the move from polished, scripted rhetoric to informal, conversational formats across leadership, media, and society. The author draws parallels between Socrates' plainspoken defense at his trial and the modern rejection of teleprompters and scripted speeches. Leaders today are gaining trust by engaging in genuine, unscripted dialogue, reflecting societal trends toward authenticity and decentralization of power. The piece critiques traditional oratory, arguing it has been degraded by divisive rhetoric and hierarchical structures, making conversational styles more relatable and influential. This transformation, exemplified by podcasts, YouTube, and Substack, marks the decline of the "talking head" era and challenges leaders to adapt or risk losing relevance.

We are past peak Woke?

Don’t idolise Roger Scruton - The Critic

• Roger Scruton’s enduring appeal lies in his humane and accessible voice, his celebration of beauty, and his philosophical focus on tradition, morality, and aesthetics. His work has offered conservatives intellectual and cultural tools, countering the modern right's philistinism and its tendency to dismiss art and intellect. Scruton’s emphasis on beauty and culture contrasts with his neglect of economic and structural questions, exposing the broader right-wing failure to address crises like inequality and the failures of both state and market.

• While Scruton championed conservatism as an instinct, critics argue this reduces it to sentimentality, overlooking richer traditions that see culture and order as grounded in divine or communal ideas. His broader insights—on craft, the welfare state, and inequality—invite conservatives to rethink their uncritical embrace of free-market ideologies. However, the veneration of Scruton risks idolizing him, limiting his legacy to a narrow, static form of conservatism, rather than using his work as a guide to a dynamic and imaginative renewal of conservative thought.

12. Miscellaneous

Why didn't people smile in old photographs? It wasn't just about the long exposure times.

People blame these serious expressions on how long they had to sit for a photo, but that's not the whole picture.

• People in 19th and early 20th-century photographs often appeared serious not only because early cameras required long exposure times—making it difficult to hold a smile—but also due to cultural norms inherited from painted portraits. Smiling was considered undignified and was associated with the poor, the lewd, or the foolish; it wasn't customary in formal portraiture to display such expressions. Additionally, the transient nature of a genuine smile was challenging to capture in static paintings, leading to a tradition of neutral or serious faces. It wasn't until the advent of motion pictures in the 1920s, which normalized capturing a range of emotions, that smiling became common in photographs, changing societal perceptions and practices around portraiture.

Scientists study 35,000-year-old saber-toothed kitten frozen in Siberia - The Washington Post

Scientists have studied the frozen remains of a 35,000-year-old saber-toothed kitten for the first time and found “significant differences” compared to a modern-day lion cub.

Two Sea Creatures Fused Into a Single Animal - The WSJ

In a strange episode in the animal kingdom, a captive marine animal merged with another of its kind to become a single individual. The organisms that merged are comb jellies, translucent invertebrates that resemble jellyfish but belong to a different group of marine animals called ctenophores.

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

I must give credit to a broad range of writers for pointing out many of these links, in particular Tyler Cowen of Marginal Revolution (and George Mason Univeristy)

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: ⁠⁠⁠⁠⁠⁠https://substack.com/@thespecialistgeneralist⁠⁠⁠⁠⁠⁠. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #8: Explore Anil Varanasi's visionary take on innovation, design, and storytelling, discover the rise of AI-powered tools like bird-identifying binoculars and sneaker-authenticating sensors, the surprising advantages of neurodivergence in elite sports, and the sustainability promise of wooden satellites. Delve into debates on the future of prediction markets, Meta's AI military partnerships, and the cultural dynamics of online gatekeeping versus recommendation culture. And, much more…

Quotations in italics

Table of contents:

1. Top things I’ve read this week

2. AI/Technology news of the week

3. Interesting AI use cases of the week

4. A Start-Up that I like

5. Science news of the week

6. World news of the week

7. USA news of the week

8. China news of the week

9. India news of the week

10. UK news of the week

11. Academic paper(s) of the week

12. Culture of the week

13. Miscellaneous

1. Top things I've read this week

EP 89 Understanding Anil Varanasi : reinventing networking, spotting talent, design and AI and a surprise episode takeover - The Aarthi and Sriram Show

• This podcast features Anil Varanasi, co-founder of Meter, in a wide-ranging and deeply insightful conversation that explores his unconventional approach to entrepreneurship, design, and talent discovery. Anil discusses his experiences growing up in an unstructured yet intellectually stimulating environment, his ventures in film and technology, and the lessons learned while building a vertically integrated networking company that emphasizes user experience and design. His belief in empowering young, undiscovered talent has led to transformative investments, exemplified by his support of figures like Dwarkesh Patel. Anil also shares his reflections on cultural shifts, long-term ambition, and the role of storytelling in shaping societal values. The episode concludes with thought-provoking questions on the future of technology, immigration, and legacy-building, showcasing Anil’s profound impact on both personal and professional levels.

• I’ll call out a few quotes.

• On China (& Shenzhen)

  • I tend to believe like people in China in individual sense are just incredible. In aggregate I think, there's a lot of issues geopolitically, like you mentioned, culturally, if you went anywhere else in 2014 or 2015, rest of the world, everybody wanted to be America, right? In the way they spoke, the way they dressed, what music, what film. Everything was how to be America. But China, what it felt different was that they wanted to beat America rather than be America. And that felt like a very stark difference. They felt like they had earned the right to compete with the largest economy. And you could feel that energy and fervor. We just want to build things. We know we can build things and we can do it at scale. And in Shenzhen, that culture was ever present. You design something in the morning. And then by night, somebody would run over to your lab saying, it's produced. Do you want to test it? It was just beautiful at that time. It's cyber punk to the core. But that, Culture was seeped in very deeply in working really hard, but also I think Shenzhen at that time, because like I was mentioning it, it was young people. They were all working really hard to send money back to families that lived elsewhere in China. I think that was also critical thing. I think India now has that to talking to folks in Africa. It feels like that fervor is there. I haven't been to China since the beginning of the pandemic, so I don't know how it's changed in the last four years, but it feels like other places have that still.

• On the future of software:

  • All it will take is to just understand a person's preferences and then software is just made for them entirely on the fly for whatever they're trying to do. And that could be any sort of thing underneath that could be CRM to HR could be finances anything. I think it will matter more on the user rather than the system of record, which is what it's been the last few years the web and we plan on pushing this too, which is even with command, what we want to do is actually understand the user deeply and put that into context of the models and other things to really take it further. But I think that will happen really at a large scale that personalization of software. will actually happen, which we all tried in the early 2000s, that just didn't go anywhere.

• On talent

  • Anil: It's very rudimentary as it should be. So my brother and I have a very simple rule the last decade. If we come across something good on the internet that might be a YouTube channel, a blog post, podcast, gitHub repo, whatever the case might be. It doesn't really matter how somebody is creating it. And it ties back to why I care about the internet, which is you get to have access into all the nodes in the world very easily compared to before when you didn't. So if we come across this and we think like it's sufficiently interesting work and there's some spark of genius there and they don't have an audience at all they have low number of subscribers or something.

  • You guys are having me spill all my secrets, but we just send them an email. That's very simple, which is how much money do you need to do this full time for six months? And that's the only question we ask

• On recommendations for young people:

  • Probably the highest thing is not waiting. The latest version of this meme is you can just do things and I think it applies the most to young people than anybody else. The downside risk is actually fairly low. You can just go back, do other things later on. And if I had to push people's ambitions is maybe take themselves more seriously than they do, because I think young people can actually have a big change…. [also] the older literature, when you go read, whether it's like novels or biographies or other things, really pushes the idea that if you're young, you go do something and you have to get on with your life.

• On film:

  • Film is one of the greatest mediums we've ever come across to actually have influence on people. I think especially India is so influenced by film. You could point to any state and have politicians that used to be in film. For a particular reason. And that's true in India compared to anywhere else at a different order of magnitude. So I think film is one way to do it.

  • Sriram: Should wealthy Silicon Valley founders and CEOs just be producing movies?

  • Aarthi: Should be in storytelling at least in some form

• On London & a sense of history:

  • I want to touch on something else Which is one of the things that living in london has opened me up to is a real appreciation of history . Because you live here and one of my favorite things about the city are these little blue plaques that are various buildings where there's about a couple of hundred of them, I believe, and it tells you this famous person lived here like 100 years ago, right?

  • Aarthi: And and it is very cool. It's very cool. Like we walk by this one pretty often, which is T. S. Eliot. And it's amazing to me that T. S. Eliot used to live here not so long ago. In this particular house, like I can imagine like him eating breakfast and writing things and being like maybe I should publish this. But it's incredible to see that.

Letter #232: Jawed Karim (2006) - Cofounder of YouTube and YVentures - From Concept to Hyper Growth

Two weeks after Google announced it was acquiring YouTube, YouTube cofounder Jawed Karim gave a talk at UIUC sharing YouTube journey from pre-inception to acquisition, aptly titled YouTube: From Concept to Hypergrowth.

In this talk, Jawed shares the process of how YouTube the product came to be, because when he sees an interesting product, the first thing he thinks about is how they came up with the idea, how they built it, what troubles they ran into building it, and why such a product hadn’t been built earlier. He starts by sharing his favorite video, then some high level stats for the platform. Next, he discusses the concept of a killer application, before breaking down several examples products that influenced/developed the base that made YouTube possible (LiveJournal, HotOrNot, Wikipedia, Friendster, del.iocio.us, Flickr). In the next section of his presentation, Jawed gives the students context on the state of the video sharing market before YouTube, then shares why the timing was right for a product like YouTube to come into existence. After sharing YouTube’s product journey, Jawed moves on to discuss its business journey, from ideation to prototype to first launch to marketing to revamp to initial traction to hypergrowth. He ends the talk by answering, or rather asking, What’s next??

• I was particularly interested by the section where Jawad reflects on the platform's inception, rapid growth, and the factors that made its success possible. He emphasises the concept of a "killer application"—an innovation that proves the usefulness of underlying technology—and traces a lineage of such apps in the social content space, including LiveJournal, HotOrNot, Wikipedia, Friendster, Delicious, and Flickr. These platforms demonstrated the viability of user-generated content, social networking, and community engagement, setting the stage for YouTube.

2. AI/Technology news of the week

Meta Permits Its A.I. Models to Be Used for U.S. Military Purposes - The New York Times

Meta will allow U.S. government agencies and contractors working on national security to use its artificial intelligence models for military purposes, the company said on Monday, in a shift from its policy that prohibited the use of its technology for such efforts. Meta said that it would make its A.I. models, called Llama, available to federal agencies and that it was working with defense contractors such as Lockheed Martin and Booz Allen as well as defense-focused tech companies including Palantir and Anduril. The Llama models are “open source,” which means the technology can be freely copied and distributed by other developers, companies and governments.

Palantir to partner to bring Claude models to classified environments for US government intelligence and defence operations

• Nabeel Qureshi: Imagine telling the safety-concerned, effective altruist founders of Anthropic in 2021 that a mere three years after founding the company, they'd be signing partnerships to deploy their ~AGI model straight to the military frontlines

Magentic-One: A Generalist Multi-Agent System for Solving Complex Tasks - Microsoft

• On Wednesday, Microsoft revealed Magnetic-One. A powerful, open-source AI system that autonomously completes complex, multi-step tasks by using multiple specialized agents. It can browse the web, retrieve information, navigate local files, write and execute code, and dynamically adjust its plans to handle errors It has a multi-agent architecture, which means one LLM can activate several agents to complete a task.

• At the center of the system is an "Orchestrator" agent that manages and coordinates four specialized agents: WebSurfer, FileSurfer, Coder, and ComputerTerminal.

• Each agent is responsible for a specific function, including web navigation, file handling, coding, and command-line operations.

Apple Explores Push Into Smart Glasses With ‘Atlas’ User Study

Apple Inc. is exploring a push into smart glasses with an internal study of products currently on the market, setting the stage for the company to follow Meta Platforms Inc. into an increasingly popular category. 

The initiative, code-named Atlas, got underway last week and involves gathering feedback from Apple employees on smart glasses, according to people with knowledge of the matter. Additional focus groups are planned for the near future, said the people, who asked not to be identified because the work is secret. The studies are being led by Apple’s Product Systems Quality team, part of the hardware engineering division.

XPENG Unveils Iron Humanoid Robot, Already Operational in EV Factory

At its AI Technology Day, Chinese electric vehicle maker XPENG unveiled Iron, its first humanoid robot. The robot is roughly the size of a human, standing 5'8" tall and weighing 154 pounds. XPENG says Iron is already working on production lines assembling their upcoming P7+ electric vehicle.

The economics of AI - An exponential view

A recent observation by Marc Andreessen that more compute isn’t yielding proportional intelligence gains, aligns with what we’ve long known about scaling laws: we need exponentially more resources for linear gains in performance. While we haven’t hit the technical ceiling yet – and OpenAI’s upcoming Orion model might demonstrate meaningful gains – the real constraint may be profitability. Training and running these massive models costs hundreds of millions, even billions of dollars, making it a game that only the biggest tech companies can play (see the State of AI report for just how large these clusters are getting).

From prediction markets to info finance - Vitalik

⁠⁠I believe that (i) prediction markets even as they exist today are a very useful tool for the world, but furthermore (ii) prediction markets are only one example of a much larger incredibly powerful category, with potential to create better implementations of social media, science, news, governance, and other fields. I shall label this category "info finance".

One technology that I expect will turbocharge info finance in the next decade is AI (whether LLMs or some future technology). This is because many of the most interesting applications of info finance are on “micro” questions: millions of mini-markets for decisions that individually have relatively low consequence. In practice, markets with low volume often do not work effectively: it does not make sense for a sophisticated participant to spend the time to make a detailed analysis just for the sake of a few hundred dollars of profit, and many have even argued that without subsidies such markets won’t work at all because on all but the most large and sensational questions, there are not enough naive traders for sophisticated traders to take profit from. AI changes that equation completely, and means that we could potentially get reasonably high-quality info elicited even on markets with $10 of volume. Even if subsidies are required, the size of the subsidy per question becomes extremely affordable.

• In his essay, Vitalik Buterin introduces the concept of "info finance," extending the utility of prediction markets beyond mere betting to a broader framework where financial mechanisms are deliberately designed to extract valuable information. He illustrates how prediction markets can serve as effective tools for aggregating dispersed knowledge, aiding in areas like governance, scientific research, and public goods funding. By aligning incentives, these markets can distill human judgment and provide insights that traditional news sources or social media may overlook. Buterin emphasises that advancements in AI and scalable blockchain technologies now make it feasible to implement info finance on a larger scale, even for micro-level decisions. He concludes that this presents a unique opportunity to leverage financial instruments to solve trust issues and enhance information dissemination across various sectors.

How ChatGPT Brought Down an Online Education Giant - The WSJ

The online education company was for many years the go-to source for students who wanted help with their homework, or a potential tool for plagiarism. The shift to virtual learning during the pandemic sent subscriptions and its stock price to record highs.

Then came ChatGPT. Suddenly students had a free alternative to the answers Chegg spent years developing with thousands of contractors in India. Instead of “Chegging” the solution, they began canceling their subscriptions and plugging questions into chatbots.

Since ChatGPT’s launch, Chegg has lost more than half a million subscribers who pay up to $19.95 a month for prewritten answers to textbook questions and on-demand help from experts. Its stock is down 99% from early 2021, erasing some $14.5 billion of market value.

3. Interesting AI use cases of the week

I Went Birding With the World’s First AI-Powered Binoculars - Wired

Swarovski Optik’s new AX Visio binoculars use image-recognition algorithms and GPS data to discern the species of whatever bird you point them at. And they work anywhere in the world. They have a built-in camera and use computer vision software to ID over 9,000 bird species in real time. They can also identify some mammals and insects, making them possible the most useful binoculars for observing wildlife. The AX Visio are available now and are unlike anything else on the market, and that uniqueness comes with a hefty price tag: A pair will set you back €4,600, which is a little over $5,000.

AI Can Now Authenticate Sneakers by Their Smell

An AI start-up working with a major sneaker resale platform believes it has found a way to authenticate footwear and potentially other products through the chemical signatures in their scent.

• Osmo is a start-up leveraging AI and scent-detection technology to combat counterfeit goods, with a focus on sneakers. By analyzing the chemical composition of materials, adhesives, and other elements that emit distinct odors, Osmo identifies subtle differences between genuine and fake products. While sniff tests are already used by human authenticators, Osmo's AI-enhanced sensors bring unprecedented precision, achieving over 95% accuracy in trials with a sneaker resale company. However, the method faces challenges, including the need for ample authentic samples to train the AI, significant investment in hardware and maintenance, and its limited applicability to items like clothing that lack strong scents.

Embed Security

• Embed Security provides an agentic security platform that is designed to take the burden off overworked security analysts by leveraging AI to autonomously investigate alerts, detect threats, and provide guidance.

• “Embed provides scalable, complete threat coverage with autonomous triaging, including the investigation and prioritization of threats across the entire security stack, while empowering the analyst, not replacing them,” the company said.

• “The agentic security platform reduces repetitive tasks and enables teams to focus on high-value work like remediation, threat hunting, and learning. Embed continuously learns and adapts to new threats, enhancing its effectiveness over time,” it added.

4. A Start-Up that I like

Naptha AI

Naptha is a modular AI platform for autonomous agents that enables developers to build AI applications, conduct AI research, and scale cooperative AI agents on the agentic web.

• Naptha AI recently announced a $6 million pre-seed raise, co-led by Arche Capital and Cyber Fund. Naptha AI envisions AGI as a network of billions to trillions of agents working together. Naptha AI offers developers a platform for building and deploying large systems of cooperating intelligent agents. The platform will allow for decentralized, multi-agent orchestration, in a scalable, trustless, and cooperative environment, shifting the narrative of AI from a singleton to a dynamic, interconnected system of entities. This market-based approach not only enhances performance and scalability but also opens the door to entirely new use cases across industries, such as software development, scientific research, predictive analytics, healthcare and other fields.

5. Science news of the week

Throw Out Your Black Plastic Spatula

Replacing a black plastic spatula with a steel or silicone option is an easy way to cut down on at least part of one's daily dose of hormone disruptors.

6. World news of the week

Where they are headed - Marginal Revolution

The Australian government has pledged to legislate an age limit of 16 years for social media access, with penalties for online platforms that do not comply. Albanese and the communications minister, Michelle Rowland, did not rule out the potential for social media users to have their faces subject to biometric scanning, for online platforms to verify users’ ages using a government database, or for all social media users – regardless of age – being subject to age checks, only saying it would be up to tech companies to set their own processes.

• Tyler Cowen: Keep in mind this move, if applied consistently, would eliminate anonymous postings.  It also would have to be enforced across a very large number of apps, even for Meta alone.  Should everyone’s biometrics be put into what might be China-hackable form?  And it means the government — not the parents — is deciding the proper level of social media access for children. Are the major social media critics for this?  Against it?  Or are they not so keen to say, one way or the other?

7. USA news of the week

Yes, you already know but Donald Trump won the US election. In light of this I couldn’t not provide some of my top links on the topic:

Rising in Status - Tyler Cowen

• In light of Donald Trumps victory, Tyler

  • 1. Prediction markets

  • 2. Competitive primary elections

  • 3. Elon

  • 4. French whales

  • 5. The integrity of the American electoral system

  • 6. J.D. Vance

  • 7. The word “trifecta”

  • 8. Twitter

  • 9. Podcasts

  • 10. Long podcasts

  • 11. The Amish

  • 12. Men

• I might also add UFC

On Elon Musk and the election, Ben Thompson writes:

What is fascinating is how this fundamentally transforms any attempt to evaluate the Twitter acquisition. From a business perspective it’s a massive failure, and might always be: Musk paid too much for Twitter as it was, and in the intervening years the flight of advertisers from the platform has made it worth even less. From a Musk Inc. perspective, however, X played a pivotal role in ensuring that the incoming administration will do whatever Musk needs at the exact moment that SpaceX is gaining the capabilities to actually make a trip to Mars, if only the FAA in particular will give him the freedom to do so. That alone is almost certainly worth $44 billion to Musk!

Time to Retire "Demographics is Destiny" - Richard Hannania

The Economist recently ran a poll of various countries across the world on whether they prefer Trump or Harris. Here are results.

• Hanannia critiques the conservative fixation on demographic determinism in political outcomes, particularly the belief that certain racial or ethnic groups inherently support left-wing policies. He argues that political allegiance among lower-income or under-educated groups across all demographics is often shaped by tribalism, aesthetics, and superficial cultural signals rather than by principled stances on liberty or economic policy. Trump’s appeal to both Hispanic men and third-world populations illustrates this dynamic, as they gravitate toward his populist style rather than a shared ideological foundation. The author contends that intellectual and political influence is more effectively cultivated through shaping elite ideas, which are then marketed to the public, rather than by trying to engineer a favorable demographic base. This "demographics as destiny" view, they argue, reflects a retreat from substantive intellectual engagement and a misplaced faith in the lower-class white base to uphold conservative principles without cultivating elite alignment.

Americans hate inflation more than they hate unemployment

In 2021 and 2022, Americans’ real wages suffered their biggest drop in postwar history, thanks to inflation:

Remember, real wages are a key economic input that went into the model that managed to successfully predict every state’s election result. People really don’t like it when their real wages go down! One key thing to understand is that this sudden impoverishment happened to most Americans. Rising consumer prices affect everyone, from the rich to the poor, the employed and the unemployed, the old and the young. In a recession, on the other hand, the negative effects are largely concentrated. If you’re one of the unlucky people who loses their job, you take a huge hit. But if you’re not one of those people, you probably do OK. Maybe your wages grow more slowly for a few years, and maybe you have to stress out about the increased possibility of being laid off. But in general, the harms of inflation are diffuse while the harms of unemployment are concentrated.

And then shifting gear away from the election:

US Regulator Rejects Amazon-Talen Nuclear Power Agreement - Bloomberg

In March, Amazon Web Services paid Talen $650 million for a 960-megawatt data center campus adjacent to the Susquehanna nuclear plant in Pennsylvania, and signed a long-term agreement to buy power from the plant. [Today] the top US energy regulator rejected a special deal that would have allowed an Amazon.com Inc. data center to use more power from an adjacent nuclear power plant.

Meta’s plan for nuclear-powered AI data centre thwarted by rare bees - The FT

Plans by Mark Zuckerberg’s Meta to build an AI data centre in the US that runs on nuclear power were thwarted in part because a rare species of bee was discovered on land earmarked for the project, according to people familiar with the matter.

8. China news of the week

China Hack Enabled Vast Spying on U.S. Officials, Likely Ensnaring Thousands of Contacts

• The Salt Typhoon hacking group, linked to Chinese intelligence, conducted a sophisticated cyber-espionage campaign targeting high-level U.S. officials, exploiting telecom networks to capture sensitive data such as call logs and some voice audio from compromised cellphone lines. This infiltration, which included a deep breach into Lumen Technologies' network infrastructure, showcases China’s growing mastery of cyber-espionage and its strategic use of advanced AI to analyze and map social networks of American officials. While the hacks underscore China's capability to access sensitive U.S. communication systems, they also reveal the extent to which Beijing appears undeterred by diplomatic fallout, instead focusing on amassing intelligence that could shape its geopolitical strategy.

9. India news of the week

Human Capital Accumulation in China and India in 20th Century

• Abstract: The education system of a country is instrumental in its long-run development. This paper compares the historical evolution of the education systems in the two largest emerging economies- China and India, between 1900 and 2018. We create a novel time-series data of educational statistics related to enrolment, graduates, teachers and expenditure based on historical statistical reports. China adopted a bottom-up approach in expanding its education system, compared to India’s top-down approach in terms of enrolment. While India had a head-start in modern education, it has gradually been overtaken by China- at Primary education in the 1930’s Middle/Secondary level in the 1970s and Higher/Tertiary level in the 2010s. It resulted in the lower cohort-wise average education and higher education inequality in India since 1907. Vocational education is a central component of the Chinese education system, absorbing half of the students in higher education. In India, the majority of the students pursue traditional degree courses (Bachelors, Masters etc.), with 60% in Humanities courses. Though India is known as the “land of engineers”, China produces a higher share of engineers. We conjecture that the type of human capital in China through engineering and vocational education helped develop its manufacturing sector. Utilizing micro-survey data since the 1980s, we show that education expansion has been an inequality enhancer in India. This is due to both the unequal distribution of educational attainment and higher individual returns to education in India.

10. UK news of the week

Why does it cost £100m for HS2 to protect bats?

Gavin Pearson, editor of New Civil Engineer, said: "In recent years, achieving a DCO [Development Consent Order] has been taking longer than many think it should. "Good plans have sat waiting for years, which means things like cost projections will be out of date. "If you multiply that same problem for planning processes at local authorities, the implications can be huge." He said engineering projects could cope with "complex planning rules", but there was a "growing inability of planning processes to keep pace with applications".

11. Academic paper(s) of the week

Generative AI and the Nature of Work

Abstract: Recent advances in artificial intelligence (AI) technology demonstrate considerable potential to complement human capital intensive activities. While an emerging literature documents wide-ranging productivity effects of AI, relatively little attention has been paid to how AI might change the nature of work itself. How do individuals, especially those in the knowledge economy, adjust how they work when they start using AI? Using the setting of open source software, we study individual level effects that AI has on task allocation. We exploit a natural experiment arising from the deployment of GitHub Copilot, a generative AI code completion tool for software developers. Leveraging millions of work activities over a two year period, we use a program eligibility threshold to investigate the impact of AI technology on the task allocation of software developers within a quasi-experimental regression discontinuity design. We find that having access to Copilot induces such individuals to shift task allocation towards their core work of coding activities and away from non-core project management activities. We identify two underlying mechanisms driving this shift – an increase in autonomous rather than collaborative work, and an increase in exploration activities rather than exploitation. The main effects are greater for individuals with relatively lower ability. Overall, our estimates point towards a large potential for AI to transform work processes and to potentially flatten organizational hierarchies in the knowledge economy.

Artificial Intelligence, Scientific Discovery, and Product Innovation

Abstract: This paper studies the impact of artificial intelligence on innovation, exploiting the randomized introduction of a new materials discovery technology to 1,018 scientists in the R&Dlab of a large U.S. firm. AI-assisted researchers discover 44% more materials, resulting in a 39% increase in patent filings and a 17% rise in downstream product innovation. These compounds possess more novel chemical structures and lead to more radical inventions. However, the technology has strikingly disparate effects across the productivity distribution: while the bottom third of scientists see little benefit, the output of top researchers nearly doubles. Investigating the mechanisms behind these results, I show that AI automates 57% of “idea-generation” tasks, reallocating researchers to the new task of evaluating model-produced candidate materials. Top scientists leverage their domain knowledge to prioritize promising AI suggestions, while others waste significant resources testing false positives. Together, these findings demonstrate the potential of AI-augmented research and highlight the complementarity between algorithms and expertise in the innovative process. Survey evidence reveals that these gains come at a cost, however, as 82% of scientists report reduced satisfaction with their work due to decreased creativity and skill underutilization.

• Caleb Watney has described this as the best paper written so far about the impact of AI on scientific discovery

• How scientists changed their views on AI before and after the experiment

Exposure to sugar rationing in the first 1000 days of life protected against chronic disease - Science

Abstract: We examined the impact of sugar exposure within 1000 days since conception on diabetes and hypertension, leveraging quasi-experimental variation from the end of the United Kingdom’s sugar rationing in September 1953. Rationing restricted sugar intake to levels within current dietary guidelines, yet consumption nearly doubled immediately post-rationing. Using an event study design with UK Biobank data comparing adults conceived just before or after rationing ended, we found that early-life rationing reduced diabetes and hypertension risk by about 35% and 20%, respectively, and delayed disease onset by 4 and 2 years. Protection was evident with in-utero exposure and increased with postnatal sugar restriction, especially after six months when solid foods likely began. In-utero sugar rationing alone accounted for about one third of the risk reduction.

• Pregnant women might want to ration their sugar intake, as well as alcohol, during pregnancy.

Early Mentors for Exceptional Students - Ian Calaway

Abstract: Although we are acquainted anecdotally with extraordinary people like Mozart and Marie Curie, there is little systematic research on how children with exceptional ability develop into truly extraordinary talents. Is the supply of extraordinary talent inelastic, dependent on a rare combination of innate gifts and the availability of mentors who are themselves world-class (Irene Joliot-Curie and her mother Marie)? Or, could the supply be fairly elastic because mentors need only have abilities within the normal range? I analyze these questions in the context of mathematics, where there is a consensus on how exceptional ability presents itself in children. I show that mathematics teachers who organize clubs and competitions can identify and foster exceptional math students, causing them to win honors, attend selective universities, major in STEM fields, and have careers in which they disproportionately spur economic growth. I demonstrate that there are many exceptional math students without mentors who could be reached with modest investments.

• Alex Tabbarok has more on this

  • Calaway estimates that over 27 years, math mentors identified 9,092 American Math Competitions students (the cream of the crop) but there were 11,168 missing students of very high ability.

    These 11,168 additional students represent the missing exceptional math talents who would have participated in the AMC and been identified as exceptional if they had access to a mentor…these mentors would have increased the number of these students attending selective universities (3,017 students), majoring in STEM (3,465 students), earning PhDs (1,652 students), and pursuing careers as scientists and professors (1,850 students) during this twenty-seven year period.

    11,168 missing students of very high ability over 27 years may not sound like much but we are talking about the very top talent level. A footnote illustrates:

    Sergey Brin (Google), Mark Zuckerberg (Meta), Peter Thiel (PayPal), and Sam Altman (OpenAI) were all top AMC scorers (Committee on the American Mathematics Competitions, 1980–2023)

How Do Business Owners Run Governments? Evidence from Brazilian Municipalities, Job Market Paper.

Abstract: Effective government leadership demands skills in planning, budgeting, and personnel management, areas where business experience may be valuable. Business owners could apply their management skills in public administration to optimize resources, enhance public service delivery, or promote private sector growth. However, these skills may not translate to the public sector, where different rules prevail, and a lack of political expertise could ultimately undermine the delivery of public services. This paper studies the impact of electing business owners as mayors in Brazilian municipalities. By leveraging a national business registry from the Brazilian Federal Revenue Agency that includes information on the owners, merged with candidates’ data from the Electoral High Court, I am able to identify business candidates. Using a regression discontinuity design for close elections, I find that electing a business mayor reduces municipal revenue by 7.5% and spending by 8%. Evidence suggests that the reduction in revenue may be due to challenges in accessing intergovernmental transfers, indicating that managerial experience may come at the expense of political skills. Despite the decrease in spending, business mayors do not negatively affect public service delivery or local economic activity. I find no impact on health outcomes and a small but positive effect on private sector employment. These results indicate improvement in cost-effectiveness by achieving equivalent service levels with fewer resources. I examine management as a driver of increased effectiveness and find that business mayors do not improve personnel management outcomes, suggesting a limited ability to transfer or adapt their management experience to the public sector.

• Interesting in light of D.OG.E and Elon and Vivek!

12. Culture news of the week

The Banality of Online Recommendation Culture | The New Yorker

⁠“In a world of scarcity, we treasure tools. In a world of abundance, we treasure taste,” Atluru wrote. Given that the Internet offers us so many options, the choice of what to pay attention to, what to consume, or even what to create matters most. By sharing your taste online, you can develop cultural capital. As Bainbridge put it, “Making the right recommendation comes with clout.”

Another buzzword—“gatekeeping”—has taken on a new and different valence online lately, to express a desire not to recommend. “Gatekeeping” means keeping insider information to yourself instead of tossing it to the winds of the Internet. In another much-discussed recent essay, the designer and artist Ruby Justice Thelot praised the gatekeeper for erecting “the fence which the enthusiast happily hops” but which “stops the dilettante”—in other words, for making it hard to experience whatever is being recommended without a bit of investment. Mundane things are easy fodder for recommendations; what is truly closest to your heart might warrant a bit of withholding, however antithetical it seems to the pressures of being online.

• The rise of human-curated recommendations, exemplified by platforms like Perfectly Imperfect, reflects both a reaction to and an evolution of algorithm-driven content. This trend emphasizes personal taste and authenticity, distinguishing itself from automated feeds by showcasing real individuals' preferences, often mundane or niche. As algorithmic suggestions dominate, curated recommendations become a form of cultural capital, with taste positioned as a valuable commodity in an era of information overload. However, this model faces challenges: the repetition of widely shared recommendations risks diluting individuality, while a growing counter-movement of "gatekeeping" values withholding personal passions to protect their uniqueness. Ultimately, the tension between efficiency, individuality, and authenticity defines this shift in digital content.

Social media basically brought us to something like an oral culture - Katherine Dee on X

- We're not illiterate in the sense that we can't read, our writing approximates our speaking
- Our writing isn't as complex because it doesn't function like text used to
- We both archive everything and trust our collective memory -- everything is saved, bookmarked, etc. but never revisited (when was the last time you bookmarked a website or even checked your own likes?)
- For information to be remembered it has to be recirculated, repeated, or go viral or we forget because time moves so fast (similar to storytelling?)
- You can't look things up easily because we live in a perpetual now -- if you don't understand the context of the discourse, you need to ask someone to catch you up
- This also makes society very participatory
- This has weird knock-on effects like needing to *always* be online to know what's going on in the world - you can't just hermit away and study, at a minimum you're lurking
- Because of the way the Internet treats time, when we do revisit the past, it's marked by events vs. actual moments in time - do you remember 2017 or what was meme-worthy then?
- There's always a current thing
- Language evolves quickly; things feel dated quickly
- We determine if things are truthfulness through vibes/tribal consensus - there are authorities but they're cults of personality vs. institutions
- You can't "just learn" things, you need to be in the right networks
- Memes, copypastas, etc. are very similar to oral cultures
- I actually don't know if any of this is right so feel free to correct me

13. Miscellaneous

Hyperfocus of ADHD and autism can be real assets for elite athletes - The Washington Post

“People with ADHD and autism have an incredibly high ability to focus on tasks that they find interesting or stimulating,” said Laura Huckins, an associate professor of psychiatry at the Yale Center for Genomic Health. “They tend to be drawn towards professions that require or include novelty, that include regular challenges, and that require high performance under stress and pressure.”

• This article explores the unique advantages and challenges neurodivergent athletes face in sports, focusing on individuals like Caragh McMurtry, an autistic Olympic rower. Neurodivergent traits such as hyperfocus, sensory sensitivity, and a high tolerance for physical repetition can enhance performance, while rigid sports structures often clash with their needs, leading to premature burnout. Studies suggest that ADHD and autism traits are prevalent among elite athletes, offering evolutionary advantages like impulsivity and risk-taking in competitive sports. However, traditional coaching methods often fail to accommodate these differences, as McMurtry experienced when she retired due to unsupportive training protocols. She and her husband, also neurodivergent, now advocate for neuro-inclusive sports environments through their initiative, Neurodiverse Sport. By embracing neurodivergent strengths and adapting systems, sports can unlock the full potential of diverse athletes while mitigating the associated challenges.

World's 1st wooden satellite arrives at ISS for key orbital test

The world's first wooden satellite, a tiny Japanese spacecraft called LignoSat, arrived at the International Space Station (ISS) today (Nov. 5) aboard a SpaceX Dragon cargo capsule. LignoSat measures just 4 inches (10 centimeters) on each side, but it could end up having a big impact on spaceflight and exploration down the road. "While some of you might think that wood in space seems a little counterintuitive, researchers hope this investigation demonstrates that a wooden satellite can be more sustainable and less polluting for the environment than conventional satellites," Meghan Everett, the deputy chief scientist for NASA's International Space Station program, said in a press briefing on Monday (Nov. 4), a few hours before the Dragon capsule lifted off.

World war ant has been raging for over a century, from Japan to South Africa.

• Read it - you won’t regret it - there is indeed a world war ant.

The Martian Aerogel Jacket

A 700 gram jacket that combines aerospace-grade materials and cutting-edge insulation technology to bring NASA-level protection to outdoor wear. Pretty cool!

During the late warring states/Han dynasty people would be buried with official ID to get them into the underworld

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

I must give credit to a broad range of writers for pointing out many of these links, in particular Tyler Cowen of Marginal Revolution (and George Mason Univeristy)

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: ⁠⁠⁠⁠⁠https://substack.com/@thespecialistgeneralist⁠⁠⁠⁠⁠. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #7: Explore the current state of artificial intelligence (AI), its impact on various sectors, and the challenges it presents. Examine AI's role in advertising, content creation, and search, as well as its influence on the business landscape, including the rise of AI-powered companies and the potential for AI-driven job displacement. Delve further into the potential of AI in the military, healthcare, and climate change, while addressing global news such as the ongoing competition between the US and China in AI development and the growing concern over nuclear risk.

Quotations are in italics

Table of contents:

1. Top things I’ve read this week

2. AI/Technology news of the week

3. Interesting AI use cases of the week

4. Start-Up news of the week

5. Science news of the week

6. Climate news of the week

7. Sports business news of the week

8. USA news of the week

9. World news of the week

10. China news of the week

11. Military news of the week

12. Academic paper(s) of the week

13. Culture of the week

14. Miscellaneous

1. Top things I've read this week

Meta’s AI Abundance - Stratechery

• Ben presents a comprehensive, bullish case for Meta as the most strategically positioned company to capitalise on generative AI and its implications across the short-, medium-, and long-term.

Short-Term:

• Generative AI's Impact on Digital Advertising

  • Meta already uses advanced machine learning to power its ad system, offering advertisers automated, outcome-based solutions. This infrastructure, combined with generative AI, gives Meta a clear competitive advantage.

  • Generative AI allows advertisers to provide basic brand guidelines and broad parameters, after which Meta can generate and test countless variations of ads. Meta's system will be able to learn from the success of each ad in real time, improving its ability to target and convert audiences. This dynamic creates a feedback loop that competitors with less robust data and infrastructure cannot replicate.

• Black Box Advertising Model

  • Margin Expansion: Meta’s ad model functions like a black box, optimizing ad placement and pricing to maximize advertiser results while increasing its own margins. Generative AI amplifies this effect.

  • ATT Aftermath: Apple’s App Tracking Transparency (ATT) disrupted deterministic ad targeting across the industry, but Meta adapted by developing probabilistic models. This required massive investment in GPUs and AI infrastructure, which Meta leveraged effectively, outpacing competitors who couldn’t meet the new data and processing demands.

• Text-to-Message Ads

  • AI Agents for Business Interactions: Meta is developing AI-powered agents that handle customer interactions via messaging, addressing a key labor cost issue for e-commerce, particularly in regions with higher labor costs.

  • This opens up new monetization channels while reinforcing Meta’s dominance in messaging markets.

• Strategic GPU Investment:

  • Meta’s significant GPU purchases and infrastructure investments position it uniquely to generate short-term returns from generative AI, compared to speculative bets by competitors like Microsoft, Google, and Amazon.

Medium-Term: Infinite Content and the Smiling Curve

• AI Content Generation

  • Infinite Content: Meta is integrating generative AI directly into user feeds, producing personalized content based on user preferences. This has the potential to enhance engagement and increase ad inventory.

    • Cost vs. Value: While generating AI-driven content incurs higher costs than user-generated content, the potential for increased engagement and ad conversions justifies the expense.

    • Blurred Lines Between Ads and Content: Ads and content become indistinguishable in this paradigm. For instance, personalized generative content might subtly include products or links to purchases, seamlessly blending ad inventory with organic content.

      • You can already see the outlines of that given I’ve discussed both generative ads and generative content; they’re the same thing! That image that is personalized to you just might happen to include a sweater or a belt that Meta knows you probably want; simply click-to-buy. It’s not just generative content, though: AI can figure out what is in other content, including authentic photos and videos. Suddenly every item in that influencer photo can be labeled and linked — provided the supplier bought into the black box, of course — making not just every piece of generative AI a potential ad, but every piece of content period.

• AI-Powered Search and Discovery

  • Meta’s AI-driven search functionality could rival traditional search engines, creating a new revenue stream through search advertising.

  • By controlling both the content and the discovery mechanism, Meta can ensure a tighter integration of ads and organic user experiences.

• Meta's Unique Advantage

  • Long-Tail Advertising Power: Generative AI democratizes ad creation, enabling smaller advertisers to produce high-quality creative at scale. This increases demand for Meta’s ad inventory and drives prices higher over time.

  • Competitors Face Challenges: Rivals like Google and Amazon may leverage generative AI, but their ad dominance is tied to different parts of the funnel (search and transactional). Meta’s position at the top of the funnel benefits the most from generative AI’s creative and targeting capabilities.

Long-Term: Extended Reality (XR) and Generative UI

• AI's Role in XR Adoption

  • Generative UI: AI enables dynamic, on-the-fly user interfaces tailored to specific tasks, which is particularly important for XR platforms. This eliminates the need for traditional UI design, making AR and VR experiences more frictionless and intuitive.

  • Hardware Integration: Meta’s Orion neural wristbands and Quest headsets showcase its commitment to marrying AI advancements with cutting-edge hardware.

• The Metaverse Vision

  • Generative Worlds: AI reduces the cost and effort required to create immersive 3D environments, overcoming one of the main obstacles to metaverse adoption. This aligns with Meta’s broader metaverse strategy, leveraging its AI expertise to lead in virtual worlds.

  • Advertising in XR: Generative AI enables every element in the metaverse to serve as potential ad inventory, creating new monetization opportunities.

What Do You Do With an Idea? - Packy McCormick

⁠⁠Why are people building things again now? Here’s my hypothesis:

Ideas have gotten more expensive not to discover, but to build, at the same time that demand for premium versions of things slowed their spending. That meant that the initial push down the cost curve never happened, which meant that ideas remained more expensive to implement, which means they never got to the normal part of the demand curve.

Now, two things are happening.

1. Technology is making old and new ideas cheaper and easier to build.

2. There’s a reemergence of “premium demand.”

Ideas Need Adoption to Make an Impact. Ideas alone are not enough. They need to be techno-economically viable, buildable for a low enough cost that the market is able to adopt them. There is no progress without adoption

• The article argues that the apparent decline in new ideas isn’t due to a lack of innovation but rather the growing difficulty in implementing and scaling existing ideas—especially older ones from the mid-20th century. Despite increasing research input, regulatory and economic barriers have made it harder to translate ideas into progress, which contributes to a stagnation in Total Factor Productivity (TFP). Now, however, two trends are revitalising innovation: advanced technologies that reduce implementation costs and a resurgence of "premium demand" from sectors like Big Tech and defense, willing to pay more for capabilities that only cutting-edge ideas can provide. These shifts suggest that with reduced regulatory constraints and a focus on practical execution, we could unlock a backlog of viable ideas, propelling a new era of technological and economic growth.

The Network Society Docuseries - Ep1: The World's First Startup City

• This narrative explores the concept of decentralisation as a transformative solution to systemic governance failures, focusing on examples like Próspera in Honduras and the broader potential of Network Societies. By leveraging decentralised technologies and governance models, these initiatives aim to reduce inefficiency, foster economic freedom, and spur innovation in areas such as finance, healthcare, and urban planning. Drawing parallels to startups, these experiments prototype new systems at small scales, enabling adaptability and learning. While challenges remain, including resistance from entrenched systems, the vision posits that decentralisation, technological progress, and collective experimentation could create a more prosperous, equitable, and resilient future.

2. AI/Technology news of the week

Introducing ChatGPT search - OpenAI

ChatGPT can now search the web in a much better way than before. You can get fast, timely answers with links to relevant web sources, which you would have previously needed to go to a search engine for. This blends the benefits of a natural language interface with the value of up-to-date sports scores, news, stock quotes, and more.

• ChatGPT will choose to search the web based on what you ask, or you can manually choose to search by clicking the new web search icon.

• AMA with OpenAI’s Sam Altman, Kevin Weil, Srinivas Narayanan, and Mark Chen - Reddit

  • I thought these two responses were important to note. Agents! And what should new grads be focused on!

Google is reportedly developing a ‘computer-using agent’ AI system

Google could preview its own take on Rabbit’s large action model concept as soon as December, reports The Information. “Project Jarvis,” as it’s reportedly codenamed, would carry tasks out for users, including “gathering research, purchasing a product, or booking a flight,” according to three people the outlet spoke with who have direct knowledge of the project.

Powered by a future version of Google’s Gemini, Jarvis reportedly only works with a web browser (it’s tuned specifically for Chrome). The tool is aimed at helping people “automate everyday, web-based tasks” by taking and interpreting screenshots and then clicking buttons or entering text, The Information writes. In its current state, it apparently takes “a few seconds” between actions.

Tools, Agents, Interfaces, and Synthetic Luck: AI Interaction Models - The Diff

You're an early adopter of AI, whether you like it or not, because AI is being deployed in more or less visible ways in all sorts of products that you use. And we're early. But watching this transition means staying aware of the four ways end users actually interact with AI, all of which are somewhat overlapping.

• The taxonomy:

  • Tools: AI tools like chatbots, code completion, and generative media services operate within a self-contained scope. They use external data—mostly human-generated or from human-designed processes—to fulfill user requests. Since they work in isolation, using AI solely as tools poses less existential risk.

  • Agents: Agents are more complex than basic AI tools. While all AI tools interpret user requests, agents operate across multiple contexts and continue working until a task is completed or abandoned. Their ability to iterate and function autonomously in various environments sets them apart.

  • Interfaces: AI can serve as an interface that simplifies complex tasks. When you know your desired outcome but not how to start, AI interfaces determine which tools to use—whether simple ones like email or specialized ones like domain-specific AI—to achieve your goal. They bridge the gap between your request and the tools needed to fulfill it.

  • Synthetic Luck: AI creates "synthetic luck" by reducing randomness in software interactions. It enhances recommendations and predictions, making experiences feel timely and personalized—like seeing relevant emails, ads, or content just when you need them. This improved personalization increases user engagement and provides more data to further refine AI models.

• How this impacts us:

  • Many people are using all these AI applications, but they require different skill levels. For tools, you learn over time what's best done manually, what to Google, and where AI like Claude or ChatGPT can help. Agents will transform knowledge workers into team leaders managing mostly AI assistants. Those who make decisions executed by others are typically better compensated than those who execute decisions made by others. Embracing this shift allows you to focus on more interesting work, delegate routine tasks, and find new efficiencies.

  • As we progress from tools to agents to interfaces to synthetic luck, there's a qualitative change. AI tools involve you directing the AI and deciding which mental tasks to automate. Beyond that, you defer more to algorithms until AI invisibly shapes the data inputs you receive without your explicit request. This can be valuable since recommendation algorithms need to be effective to succeed. However, while users of tools and agents actively consider the roles of humans and AI, those relying on AI interfaces or content recommendations have outsourced this introspection. The designers optimizing your media consumption focus on user retention for revenue, not necessarily your long-term interests.

  • Relying increasingly on powerful, opaque AI tools isn't the best way to align our habits with our goals. In AI software, there's a tight feedback loop between user behavior and the tools offered. As these tools become more advanced, they serve as an interface between a few who use them effectively and many whose behavior is predicted and guided by the algorithms they've helped train.

Tech Giants Are Set to Spend $200 Billion This Year Chasing AI - Bloomberg

The capital expenditures of the four largest internet and software companies — Amazon.com Inc., Microsoft Corp., Meta Platforms Inc. and Alphabet Inc. — are set to total well over $200 billion this year, a record sum for the profligate collective. Executives from each company warned investors this week that their splurge will continue next year, or even ramp up.

How to Build the Future of AI in the United States By Tim Fist and Arnab Datta

Here is Azeem Azhar:

• The US might face a 100GW shortfall in electricity for AI data centres in the US – equivalent to 100 nuclear power plants. While this estimate overstates the gap by ignoring expected efficiency gains, the challenge is real. The demands of AI (and the far larger demands of the electrification of transport, heating and industry) will need a different type of grid – one which more closely resembles the two-way internet. We are in the midst of a systematic shift in our energy system. Energy is becoming technology, not a commodity. Applying new thinking (and new reformed regulation) will open up the power market…….

Researchers say an AI-powered transcription tool used in hospitals invents things no one ever said

OpenAI has touted its artificial intelligence-powered transcription tool Whisper as having near “human level robustness and accuracy.” But Whisper has a major flaw: It is prone to making up chunks of text or even entire sentences, according to interviews with more than a dozen software engineers, developers and academic researchers. Experts said that such fabrications are problematic because Whisper is being used in a slew of industries worldwide to translate and transcribe interviews, generate text in popular consumer technologies and create subtitles for videos

More than a quarter of new code at Google is generated by AI - The Verge

“More than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers,” CEO Sundar Pichai said on the company’s third quarter 2024 earnings call.

Universal Music Strikes Strategic Deal With “Ethical AI Music Company” Klay Vision - The Hollywood Reporter

Michael Nash, executive vp and chief digital officer of Universal Music Group (UMG) said the music giant was excited “to explore new opportunities and ethical solutions for artists and the wider music ecosystem, advancing generative AI technology in ways that are both respectful of copyright and have the potential to profoundly impact human creativity. UMG has always endeavoured to lead the music industry in driving innovation, embracing new technologies and supporting entrepreneurship while protecting human artistry.”

Elon Musk’s xAI in Talks to Raise Funding Valuing It at $40 Billion - The WSJ

Wall Street Giants to Make $50 Billion Bet on AI and Power generation - The WSJ

KKR and Energy Capital Partners have agreed to invest a combined $50 billion in data-center and power-generation projects to support the development of artificial intelligence.

The investment is a bet on AI’s huge energy needs and the mounting stress it is putting on the U.S. power grid. Much of it will be invested over the next four years, the companies said.

AI is killing remote work- Sahil Lavingia on X

With AI handling much of the execution work - writing code, generating content, creating designs - the main bottlenecks are now cognitive: getting stuck on problems, running low on energy, or struggling to generate fresh ideas. In-person collaboration is particularly powerful for overcoming these barriers. The spontaneous discussions, quick whiteboarding sessions, and energy of working together help teams think better, learn faster, and get unstuck more quickly.

The primary advantage of remote work in the AI era may be the ability to maintain 24/7 operations through distributed teams. While an in-person team can ship 8 hours a day, a globally distributed team can ship 24 hours a day. 3x more!

However, this works best as a complement to, rather than a replacement for, a strong in-office presence. The reason is less about speed and more about velocity.

OpenAI builds first chip with Broadcom and TSMC, scales back foundry ambition

OpenAI is working with Broadcom (AVGO.O) and TSMC (2330.TW) to build its first in-house chip designed to support its artificial intelligence systems, while adding AMD (AMD.O) chips alongside Nvidia (NVDA.O) chips to meet its surging infrastructure demands, sources told Reuters.

OpenAI, the fast-growing company behind ChatGPT, has examined a range of options to diversify chip supply and reduce costs. OpenAI considered building everything in-house and raising capital for an expensive plan to build a network of factories known as "foundries" for chip manufacturing.

The company has dropped the ambitious foundry plans for now due to the costs and time needed to build a network, and plans instead to focus on in-house chip design efforts, according to sources, who requested anonymity as they were not authorized to discuss private matters.

Inside the World's Largest AI Supercluster xAI Colossus

• This video dives into the groundbreaking development of the world's largest AI training cluster, spearheaded by XAI and powered by over 100,000 GPUs and cutting-edge infrastructure from Supermicro. Built in just 122 days, the facility represents a monumental engineering feat, boasting liquid-cooled systems, ultra-fast 400 GB Ethernet networking, and Tesla Mega Packs to stabilize power surges during intensive training jobs. Designed to support advanced AI models like Grok, the cluster integrates modular, serviceable racks with state-of-the-art cooling systems and massive centralized storage. This facility not only demonstrates rapid innovation in AI hardware but also sets a new standard for scaling AI infrastructure, paving the way for future advancements in AI applications and research.

HOVER: Versatile Neural Whole-Body Controller for Humanoid Robots

Humanoid whole-body control requires adapting to diverse tasks such as navigation, loco-manipulation, and tabletop manipulation, each demanding a different mode of control. For example, navigation relies on root velocity tracking, while tabletop manipulation prioritizes upper-body joint angle tracking. Existing approaches typically train individual policies tailored to a specific command space, limiting their transferability across modes. We present the key insight that full-body kinematic motion imitation can serve as a common abstraction for all these tasks and provide general-purpose motor skills for learning multiple modes of whole-body control. Building on this, we propose HOVER (Humanoid Versatile Controller), a multi-mode policy distillation framework that consolidates diverse control modes into a unified policy. HOVER enables seamless transitions between control modes while preserving the distinct advantages of each, offering a robust and scalable solution for humanoid control across a wide range of modes. By eliminating the need for policy retraining for each control mode, our approach improves efficiency and flexibility for future humanoid applications.

3. Interesting AI use cases of the week

Video-to-Music Generation with Semantic Alignment and Rhythmic Synchronization

Abstract. Generating music that aligns with the visual content of a video has been a challenging task, as it requires a deep understanding of visual semantics and involves generating music whose melody, rhythm, and dynamics harmonize with the visual narratives. This paper presents MuVi, a novel framework that effectively addresses these challenges to enhance the cohesion and immersive experience of audio-visual content. MuVi analyzes video content through a specially designed visual adaptor to extract contextually and temporally relevant features. These features are used to generate music that not only matches the video’s mood and theme but also its rhythm and pacing. We also introduce a contrastive music-visual pre-training scheme to ensure synchronization, based on the periodicity nature of music phrases. In addition, we demonstrate that our flow-matching-based music generator has in-context learning ability, allowing us to control the style and genre of the generated music. Experimental results show that MuVi demonstrates superior performance in both audio quality and temporal synchronization.

AI-powered Michael Parkinson podcast to launch, working with family estate

Deep Fusion Films has announced the commission of a new 8-part podcast series Virtually Parkinson hosted by an AI replica of the late Sir Michael Parkinson to be launched later this year... The podcast series will be a completely unscripted conversation between a guest and an AI version of Sir Michael Parkinson. The AI ‘Michael Parkinson’ is built using knowledge garnered from his career, including his back-catalogue of over 2000 interviews with some of the most important cultural figures of our time. This approach enables AI Michael to converse in an authentic manner and offers podcast guests a true-to-life experience.

• I suspect we are going to see much much more of this….

The 3 AI Use Cases: Gods, Interns, and Cogs

• This thoughtful exploration proposes a clear framework for categorising AI use cases, helping to demystify the AI landscape by grouping applications into three distinct categories: Gods, Interns, and Cogs

  1. Gods are advanced, autonomous AI systems aimed at replacing human intelligence in broad, complex tasks. These are still theoretical or aspirational due to technological limitations and high resource requirements. They are mostly a focus for major, well-funded entities and influence regulatory, political, and funding discussions.

  2. Interns are supervised AI tools designed to assist experts in specific fields by handling routine or repetitive tasks, enabling increased productivity without removing human oversight. These AI copilots, like GitHub Copilot or Adobe Firefly, enhance the capabilities of professionals and represent the most value-producing AI category today. Toys, a playful subset of Interns, target non-experts and are built more for entertainment than productivity.

  3. Cogs are specialized AI functions optimized for a single, well-defined task with low error tolerance, usually integrated into broader data pipelines or systems without requiring oversight. They are the unsung heroes in enterprise environments, providing stable, cost-effective AI utility within larger workflows.

Radio station in Poland fired its journalists and replaced them with AI 'presenters

Weeks after letting its journalists go, OFF Radio Krakow relaunched this week, with what it said was "the first experiment in Poland in which journalists... are virtual characters created by AI". The station in the southern city of Krakow said its three avatars are designed to reach younger listeners by speaking about cultural, art, and social issues including the concerns of LGBTQ+ people.

4. Start-Up news of the week:

The Second $100B AI Company

My prediction is this: the first AI company to eclipse $100B in market cap will be a consumer AI company… The cloud transition gave us roughly 20 companies doing over $1B in revenue; the mobile transition gave us another 20. So what will be the 20 companies borne from the AI revolution that do $1B+ in revenue?

As AI’s foundation layer has crystallized, we’ve seen a few fiefdoms emerge. OpenAI is aligned with Microsoft, of course; Anthropic is partnered with Amazon / AWS; Google has DeepMind; and Meta is investing heavily into the space. The (likely) winners at this layer of the stack are set. So if the foundation layer is stabilizing, where does new value get created? The application layer, of course.

From Sequoia’s recent piece on AI:

Imagine you want to start a business in AI. What layer of the stack do you target? Do you want to compete on infra? Good luck beating NVIDIA and the hyperscalers. Do you want to compete on the model? Good luck beating OpenAI and Mark Zuckerberg. Do you want to compete on apps? Good luck beating corporate IT and global systems integrators. Oh. Wait. That actually sounds pretty doable!

• Examples of Potential $100B Consumer AI Companies

  1. New Generative Video Platforms:

     • AI tools that let creators generate professional-grade video content could rival TikTok or YouTube by creating ecosystems where creators, influencers, and brands thrive.

  2. New AI-Native Gaming Ecosystems:

     • A platform enabling non-technical users to create immersive games—complete with generative assets, storytelling, and multiplayer environments—could create a Roblox or Minecraft successor.

  3. New AI-Driven Personalization in E-Commerce:

     • Platforms offering hyper-personalized shopping, akin to having a 24/7 personal stylist, could unlock massive growth in AI-powered commerce.

  4. New Digital Companions or Assistants:

     • The next evolution of apps like Replika or AI therapists could cater to unmet consumer needs in mental health, relationships, or productivity.

The hottest AI startups are raising back-to-back funding rounds. Suddenly, it's feeling very 2021 in Silicon Valley for a chosen few.

Inside the Secret School for the World’s Best Founders - the Generalist

• This is a really interesting deep-dive on Avra Capital, a unique force in venture capital, combining education, tactical guidance, and investment in growth-stage startups. By leveraging its YC Continuity roots, Avra focuses on equipping CEOs of breakout businesses with the skills needed to scale effectively—hiring top-tier executives, maintaining high shipping speed, and planning for public market readiness. Its curriculum, taught by experienced founders, emphasizes actionable insights delivered in candid, off-the-record sessions. Avra’s model flips the traditional VC process, proving value first through its program and building relationships before investment. With a bespoke sourcing system, a structured curriculum, and plans to expand into pre-IPO support, Avra is redefining venture capital by prioritizing founder success over transactional deals, making it both a school for scalers and a strategic investor

How Union Square Ventures Built an AI Brain for Venture Capital - Ep. 36 with Matt Cynamon

• Union Square Ventures (USV) is developing an AI-powered operating system tailored to enhance the workflow and decision-making processes of their investment team. This podcast is a great overview of how they are building and leveraging this system:

  • Centralizing Knowledge Through "The Librarian"

    • What It Does: "The Librarian" is an internal AI assistant built on top of USV's extensive history of writing and thought leadership. It uses natural language processing to allow team members to query and engage with USV’s archive of over 15,000 blog posts, memos, and meeting notes.

      • E.g. find relevant insights about potential investments by cross-referencing past writings with new opportunities. Generate tailored, high-quality responses that synthesize past opinions, lessons, and thesis-aligned perspectives.

  • AI-Enhanced Decision Support for Investments

    • USV has created GPT-based tools that can analyze and synthesize external data, such as company about pages, pitch decks, and market reports e.g. compare a pitch to relevant past investments and blog posts to evaluate alignment with USV’s investment thesis.

  • The Vision for the Future

    • Evolving "The Librarian": Turning it into a public-facing tool that could:

      • Allow entrepreneurs to query USV’s thesis before pitching.

      • Equip them with relevant writings, insights, and ideas that align with USV’s thinking.

Daniel Khachab: "We Are in the Middle of a Cold War for AI Talent" | E1220

• This is a great chat between Harry Stebbings and Daniel Khachab, co-founder and CEO of Choco, a company dedicated to streamlining and digitizing processes between restaurants and distributors. Some key takeaways are as follows:

• Transition to AI-Centric Models:

  • Daniel emphasises the need for companies to adopt AI-first strategies, where the majority of revenue or operational efficiency stems directly from AI applications. This shift is driven by: the commoditisation of traditional SaaS tools & the necessity for companies to continuously innovate or risk being disrupted by competitors leveraging AI.

• The "Agentification" of Software:

  • The future of software lies in removing traditional user interfaces, where tasks are handled by intelligent agents responding to prompts, significantly reducing the learning curve for new users.

  • Examples include payroll processing, customer service, and HR, which can all be streamlined by AI agents.

• Why SaaS is "Dead"

  • Transition from Interfaces to Intelligent Agents:

    • Traditional SaaS tools require users to interact with structured user interfaces, forms, dashboards, and workflows. These interfaces are designed to help users manage specific tasks (e.g., payroll, HR, customer service). The emerging AI-driven paradigm ("agentification") replaces these interfaces with intelligent agents. Instead of navigating through a UI, users simply issue natural-language commands, and the agent executes the task autonomously.

      • Example: Instead of opening a payroll software UI, selecting employees, and managing deductions, you could simply prompt an AI, "Process payroll for November and notify the tax authority."

  • Faster Adoption Curve:

    • AI-driven systems significantly reduce the learning curve for software. Traditional SaaS requires training users on the interface and workflows, while AI agents mimic natural human communication.

    • This ease of use is expected to make AI-based solutions more appealing, particularly in industries resistant to digital transformation, like manufacturing or traditional retail.

  • Shifting from Features to Outcomes:

    • In traditional SaaS, companies differentiate through features, functionality, and customization. AI agents, however, focus purely on delivering outcomes (e.g., processed payroll, resolved customer queries) without the intermediary step of navigating features.

    • This results in tools that are more intuitive, efficient, and productive.

  • Potential pricing shift:

    • Traditional SaaS pricing models (seat-based, feature-tiered subscriptions) may become less relevant as AI agents handle multiple roles and tasks.

    • SaaS companies need to adapt their business models to remain competitive, potentially shifting to usage-based or performance-based pricing.

  • SaaS in Transition, Not Literal Death:

    • Even in an AI-driven world, businesses need centralized "single sources of truth" like CRMs or ERPs. These data repositories will underpin the AI-powered agents that interact with them.

    • SaaS platforms like Salesforce or Workday may evolve into backend data layers, with AI agents acting as the interface.

    • SaaS companies that integrate AI capabilities will survive and thrive, while those clinging to traditional models may struggle.

    • Foundational AI companies (e.g., OpenAI, Anthropic) may blur the line between infrastructure and application layers by offering direct AI-driven solutions.

Investing in AI Native Startups - Cathy Gao of Sapphire Ventures on Bloomberg

• Cathy notes that they consider investing in companies that might have exit opportunities into big tech…

• That many companies/organisations across industries still use old computers etc. and ripe for disruption… small businesses backbone of most global economies

5. Science news of the week:

Appetite for change: GLP-1s and society - An Exponential View

We believe GLP-1 drugs will be one of the most impactful technologies in the US over the next five years (as we argued in our previous essay on GLP-1s). When used for medical purposes alone, they could affect 20% of US adults — 50 million people. But they’ll go much further. Today, we’ll lay out three scenarios we foresee could take place depending on regulation, R&D and societal acceptance. These three scenarios are: Medical, Lifestyle enhancemen, Everyone, everywhere

And more on GLP-1 from Eric Topol: Another GLP-1 success, semaglutide for relief of knee osteoarthritis in people with obesity...

Psilocybin leads to global increases in brain network integration

Overall, consistent efficacy-related brain changes, correlating with robust antidepressant effects across two studies, suggest an antidepressant mechanism for psilocybin therapy: global increases in brain network integration.

• This is a short digestible X thread commenting on a new study in @NatureMedicine which assessed the subacute impact of psilocybin on brain function in two clinical trials of depression.

Cryonics is free - Less Wrong

Cryonics is pretty much money-free now—one of the most affordable ways to dispose of your body post-mortem. Don't die because you didn't double-check whether the argument you came up against cryonics in 5 seconds checks out.

There aren't enough smart people in biology doing something boring

There are relatively few smart people in this field trying to do boring things and do them well. Put another way, everyone of note in biotech wants to be George Church (arguably one of the greatest biology researchers alive), no one wants to be Don Combs (the founder of New England Biolabs, the reagent manufacturing company fueling America’s biological research engine). But that’s not the case in software.

• This essay highlights a structural imbalance in the biotech industry, where most top-tier talent and funding gravitate toward ambitious, revolutionary projects—such as curing aging or creating groundbreaking drugs—while relatively few focus on "boring but impactful" ventures. Drawing a comparison to software, where many successful companies like Stripe, DocuSign, and Zoom thrive on perfecting seemingly mundane tasks, the author argues that biology needs a similar shift. Companies like New England Biolabs and Plasmidsaurus, which excel at providing essential services and tools for the broader ecosystem, demonstrate the untapped potential of boring-but-useful biotech initiatives. The essay also emphasizes opportunities in drug repurposing, contract research organizations (CROs), and lab automation as areas ripe for improvement. By embracing these "boring" ventures, biotech can foster more accessible, replicable, and impactful science, ultimately benefiting the ecosystem as much as or more than pursuing only grandiose ideas.

6. Climate news of the week:

The robustness of coal? - Marginal Revolution

Coal consumption in 2030 is now estimated 6% higher than only a year ago. That may sound small, but it amounts to adding the equivalent of the consumption of Japan, the world’s fourth-largest coal burner. By 2030, the IEA now believes coal consumption will remain higher than it was back in 2010…

One notable statistic: Two-thirds of the total increase in energy demand in 2023 was met by fossil fuels, according to the IEA.

7. Sport’s business news of the week:

Tapping into New Markets: How Guinness Leverages the Premier League Globally and Locally ⚽️ - The Sport’s Pundit

Guinness and the Premier League are now activated together in 88 countries, exceeding even the global reach of Guinness’ St. Patrick’s Day celebrations

• Guinness’ recent growth and brand success are closely tied to its strategic alignment with the Premier League, capitalizing on both global reach and local resonance. By becoming the "Official Beer of the Premier League" in a £40 million deal, Guinness taps into the league’s vast international viewership—1.87 billion followers weekly (34% of which are new in the last four years)—providing it with year-round exposure. Guinness’ approach leverages the league’s frequency of engagement, ensuring constant brand presence and relevance across diverse markets. This partnership, along with Guinness' pandemic-era support for pubs and innovations like Guinness Zero, underscores the brand’s commitment to “communion,” uniting people through shared social experiences, whether in a UK pub or through culturally tailored ads in places like Seoul or the Scottish Outer Hebrides. By empowering local teams to adapt brand activations to regional identities, Guinness demonstrates the importance of combining a global brand strategy with a “local heartbeat,” a model other brands could emulate to achieve both scale and cultural relevance.

YouTubers are Disrupting Sports Content - Aaron Miller at Will Ventures

To summarize, we see three major trends at the intersection of YouTube and sports:

1. YouTubers are redefining who is a sports star, and what sports content looks like

2. New sports leagues are emerging around these YouTube creators

3. New business models are emerging from YouTube sports content

• YouTubers are transforming the sports content landscape in three major ways: first, by reshaping what it means to be a sports star and reimagining sports content; second, by inspiring new leagues built around their own brands and fan communities; and third, by pioneering alternative business models, such as sponsorship-first livestreams, consumer products, and live events, which diverge from traditional sports’ reliance on media rights and ticket sales. Examples include creator-led sports leagues like the Kings League and events by personalities like Dude Perfect and Sidemen, which are tailored to engage younger audiences. YouTube’s model allows creators to build and monetize loyal fanbases through a blend of free media and diversified revenue streams, positioning them as agile competitors to legacy sports leagues that are still adjusting to changing consumption habits among younger fans. For investors, the article suggests that backing these creator-led sports ventures presents a promising growth avenue in the evolving sports and media landscape.

More than Messi: The media questions facing the FIFA Club World Cup - The Unofficial Partner

A face-saving outcome for FIFA, without putting a below-expectation market valuation on the reimagined property, might indeed be a combination of PIF underwriting revenues and reach-first distribution via FIFA+ — to go back to market with more of an audience track record for the 2029 edition.

• The expanded FIFA Club World Cup faces significant challenges in securing its financial and operational viability amid a crowded sports calendar, strained broadcaster budgets, and resistance from key stakeholders like leagues and players' unions. FIFA’s high revenue targets clash with broadcasters’ reluctance to commit to a short-format tournament without proven audience traction, further strained by competing priorities such as the Gold Cup and other domestic rights obligations. While FIFA has positioned the CWC as a marquee event, it remains a nascent concept with uncertain financial outcomes, exacerbated by player welfare concerns and league criticisms over fixture congestion. To save face, FIFA may need to rely on alternative solutions, such as strategic partnerships with PIF or leveraging FIFA+ for broader distribution, aiming to solidify the tournament’s value proposition for future editions.

8. USA news of the week

🌎 The lithium jackpot in Smackover #220

Last week, a new report found that southwestern Arkansas’ Smackover Formation region could be home to between 5-19m tons of lithium — representing 35% and 136% of current estimated US lithium resources — drilling into one of the key challenges of the energy transition: critical mineral supply.

9. World news of the week

New report on nuclear risk - Marginal Revolution

“In brief, this study is the largest systematic survey of subject matter experts on the risk posed by nuclear weapons. Through a combination of expert interviews and surveys, 110 domain experts and 41 experienced forecasters predicted the likelihood of nuclear conflict, explained the mechanisms underlying their predictions, and forecasted the impact of specific tractable policies on the likelihood of nuclear catastrophe.

Key findings include:

1. We asked experts about the probability of a nuclear catastrophe (defined as an event where nuclear weapons cause the death of at least 10 million people) by 2045, the centenary of the bombings of Hiroshima and Nagasaki. Experts assigned a median 4.5% probability of a nuclear catastrophe by 2045, while experienced forecasters put the probability at 1%.

a.       Respondents thought that a nuclear conflict between Russia and NATO/USA was the adversarial domain most likely to be the cause of a nuclear catastrophe of this scale, however risk was dispersed relatively evenly among the other adversarial domains we asked about: China/USA, North Korea/South Korea, India/Pakistan, and Israel/Iran.

2. We asked participants about their beliefs on the likely effectiveness of several policy options aimed at reducing the risk of a nuclear catastrophe. Two policies emerged as clear favorites for most participants: a crisis communications network and nuclear-armed states implementing failsafe reviews. The median expert thought that a crisis communications network would reduce the risk of a nuclear catastrophe by 25%, and failsafe reviews would reduce it by 20%.”

The Emerging Age of AI Diplomacy

In this emerging era of AI diplomacy, Washington will face similar challenges in one setting after another: it will have to control the proliferation of technologies that might have critical national security implications without kneecapping American corporations or driving potential partners into the arms of China. In their negotiations with the Gulf, U.S. policymakers should make sure that they set the right precedents

• This article explores the growing importance of AI in Saudi Arabia and the UAE's efforts to diversify their economies away from oil dependency, with significant investments in AI infrastructure and technology partnerships with American firms. While U.S. policymakers see an opportunity to counterbalance China's expanding influence in the region by fostering AI collaborations, the Gulf states are unlikely to sever their ties with China, given their significant economic interdependence and strategic hedging. The U.S. must carefully balance encouraging AI tech sales while mitigating risks, such as intellectual property theft or misuse in countries with non-transparent governance structures. Despite the Gulf’s assurances of protecting U.S. technology, American policymakers are advised to remain cautious, promoting AI development within U.S. borders while strategically navigating AI diplomacy. The broader challenge lies in preventing advanced tech proliferation without alienating potential allies or stifling American corporate interests, setting a precedent for future global AI engagements.

• UPDATE: OpenAI Says US Allies Should Partner on AI to Compete With China - Bloomberg

  • OpenAI is calling for the US and its allies to work together to support the infrastructure needed to develop artificial intelligence systems and compete with China… the US and neighbouring countries should form a “North American Compact for AI” that can streamline access to talent, financing and supply chains for building out the technology. The company said this collaboration could later expand to include a “global network of US allies and partners,” including countries in the Middle East.

  • The proposal was included in a new policy blueprint from OpenAI offering its most detailed public suggestions yet for how the US can maintain its lead in artificial intelligence and meet the technology’s significant energy demands.

    OpenAI said the US should backstop costly energy infrastructure projects by committing to purchase power from them. The company recommended the US establish “AI Economic Zones” that speed up the permitting process and help bring nuclear reactors back online. It also proposed expanding nuclear energy capacity by tapping the US Navy, which has built compact reactors to power submarines. “AI presents an unmissable opportunity to reindustrialize the US, and through that, generate the kind of broad-based economic growth that will revitalize the American Dream,” OpenAI said. “It also presents a national security imperative to protect our nation and our allies against a surging China by offering an AI shaped by democratic values, promoting individual choice and benefiting the most people possible.”

Sick man of Europe? Germany’s bosses sound alarm on staff illness

• Workers missed an average of 19.4 days because of illness in 2023, according to Techniker Krankenkasse, the country’s largest public health insurance provider….While it is notoriously difficult to compare data from country to country, Christopher Prinz, an expert on employment at the OECD, said Germany was “definitely among the higher countries” when it came to sick leave…. A study published in January by the German Association of Research-Based Pharmaceutical Companies (VFA), an industry body, found that were it not for the country’s above-average number of sick days, the German economy would have grown 0.5 per cent last year, rather than shrinking 0.3 per cent.

How to grow your princes

In a recent investigation (will open a link in Russian), the agency called Center “Dossier” has doxxed two previously secret sons of Vladimir Putin, Ivan and Vladimir (guess who’s daddy’s favorite)… What I found interesting was how these kids live. Keep in mind that they are now 9 and 5 years old. Most importantly of all, it’s complete isolation from the outside world. The people whom these kids meet are very few in number and thoroughly vetted by FGS (Russian analogue of the Secret Service). They get private teachers and tutors and they mostly live in a huge walled-in estate.

10. China news of the week

Nvidia’s AI chips are cheaper to rent in China than US - The FT

Supply of processors helps Chinese start-ups advance artificial intelligence technology despite Washington’s restrictions

• Despite U.S. export restrictions on Nvidia’s advanced A100 and H100 chips, these AI processors are widely available in China through various circumvention strategies, including smuggling and resale via foreign entities. Smaller Chinese cloud providers, unburdened by compliance concerns, offer significantly lower rates—about $6 per hour compared to $10 in the U.S.—reflecting a plentiful chip supply. Larger tech companies like Alibaba and ByteDance, prioritizing compliance, charge more but provide the reliability of regulated services. The persistence of black-market trade, facilitated by smugglers in Hong Kong, Japan, and Malaysia, demonstrates the challenges of enforcing export controls internationally. This access to powerful AI hardware undercuts Washington's efforts to curb China’s advancements in AI technology, as small cloud vendors and start-ups exploit the regulatory gaps to fuel China's AI growth.

Fab Whack-A-Mole: Chinese Companies are Evading U.S. Sanctions; Huawei Fab Network, WFE Vendors Cry Wolf, Framework for Future Controls

• This comprehensive analysis highlights the growing challenges of enforcing U.S. export controls in the face of China's rapid progress in semiconductor manufacturing, particularly in advanced AI chip production, driven by Huawei's state-backed fab network and broader domestic initiatives. While U.S. export controls have slowed China’s semiconductor advancements, loopholes, particularly through Huawei's state-backed fab network and SMIC’s strategic workarounds, undermine their effectiveness. Huawei’s rapid expansion into domestic chip production, fueled by state subsidies and evasive practices, highlights the urgency of stronger enforcement. While U.S. WFE suppliers have thrived despite restrictions, unchecked exports risk accelerating Chinese technological independence and eroding U.S. competitiveness. The analysis argues that to counter this, policymakers should tighten restrictions, expand the entity list, align with allies, and impose stricter upstream controls on critical semiconductor components. These measures are crucial to maintaining U.S. leadership in AI and semiconductor technologies, key to national security.

TikTok founder Zhang Yiming becomes China's richest man - BBC News

The surging global popularity of TikTok has seen the co-founder of its parent company, ByteDance, become China's richest person. According to a rich list produced by the Hurun Research Institute, external, Zhang Yiming is now worth $49.3bn (£38bn) - 43% more than in 2023.

11. Military news of the week

Christopher Kirchhoff on Military Innovation and the Future of War (Ep. 225) - Conversations with Tyler

An in-depth and fascinating conversation on technological transformation, defense innovation, and its implications for military strategy and national security.

• Drones and the end of man-mechanised warfare:

  • The US military has had to ask the Ukrainians to remove from the front all 31 of the M1A1 Abrams battle tanks that we gave the Ukrainians because a quarter of them had been destroyed by Russian kamikaze drones. So, not only is it the ascendancy of an era of drone warfare, but it’s probably the end of man-mechanized warfare, as well, on land. We can also talk about similarly epic changes at sea, in the air, and in space.

• Drone as enabling a battlefield management system

  • What’s really remarkable about the Ukrainian drone ecosystem is, they’re not just building simple kamikaze drones. They’re building surveillance drones that will automatically go fly, surveil the battlefield, return, upload images that can then be uploaded to killer drones that then go out. On top of this, you have people using open-source software to build fusion systems that take sensor data from EW [electronic warfare] detection systems’ live feed from drones and put them all together on a controller that an operating unit, going forward, actually uses to mount a coordinated attack using multiple drone systems. This is to say, in effect, Ukrainian start-ups have replicated the entire battle management system of the US Air Force. They have, effectively, AWACS drones, attack drones. It’s quite remarkable what they’ve done with open-source software and hardware that is pennies on the dollar compared to any system in the Western arsenal.

• COWEN: Is there a forthcoming era of assassination by drone? Do we know this? Can we prepare for it? What will that look like? Presidents never step outside?

  • KIRCHHOFF: To be honest, it’s not outside the realm of possibility. You can ask the question today, what US military installation or important government site is adequately defended against an extremely advanced, unprompted, surprise drone attack? I would guess the number is very few if not, perhaps, zero. There are people now very actively — not only at Homeland Security but in the Department of Defense — looking at how you defend against this, but it’s very hard. Drones are small. We don’t have early warning radars — in the same way that we missed the Chinese spy balloon — that are tuned to detect drones. Now people are asking questions: Can you use cell phone tower transmissions and use AI algorithms to process from that and detect flying small drones that we don’t expect to be there?

• On software progress:

  • Now we’re in a situation where a $2 trillion program of record is producing jets whose processors are slower than what we’re all carrying around in our pockets. This matters because if it turns out that 2027 AI is going to beat 2026 AI on the battlefield, then you would better be getting your AI from the commercial system of production, not from the bespoke military-industrial-complex system of production.

• Problems with defence procurement:

  • Defense procurement takes time, and venture investors want to back companies that can show demonstrable revenue growth. These two approaches are incompatible, and while investors are getting a bit better at handling this ambiguity (one which biotech VCs have dealt with just fine for decades) there is still more to be done.

12. Academic paper(s) of the week

Marine Microplastics and Infant Health

Abstract: A century of plastic usage has led to an accumulation of plastic waste in waterways and oceans. Over time, these wastes break down into particles smaller than 5 microns – or “microplastics” – which can infiltrate human biological systems. Despite decades of research into this emerging source of pollution, there is a paucity of direct evidence on the health impacts of microplastics exposure at a population scale. This paper reports the first empirical link between in-utero microplastic exposure and adverse birth outcomes. Our analysis is based on a dataset of 3 million births that occurred in coastal areas of 15 countries spanning four continents, which we merge with novel remote-sensing measurements of marine microplastic concentrations. We show that in-utero exposure to microplastics, particularly during the second and third trimesters, leads to a significant increase in the likelihood of low birth weight. A doubling of exposure increases low birth weight hazard by 0.37 per 1,000 births, which implies that over 205,000 cases per year globally can be attributed to microplastic exposure. We further show that aerosolization – whereby microplastic particles become airborne and inhalable due to seawater evaporation – is an important pathway for health impact, a challenge likely to escalate as ocean temperatures rise.

The Health and Employment Effects of Employer Vaccination Mandates

Abstract: Health care facilities considering mandating staff vaccination face a difficult tradeoff. While additional vaccination coverage will directly reduce disease transmission within the facility, the imposition of a mandate may also cause vaccine-hesitant staff to quit, which could harm patient care. To study this tradeoff, we leverage comprehensive administrative data covering virtually all US nursing homes, including payroll-based records on approximately 500 million daily nurse shifts and weekly data on COVID transmission and mortality at each facility. We use a difference-in-differences framework to estimate the impact of employer-imposed vaccine mandates at 581 nursing homes on disease spread, employment outcomes, and several patient care metrics. While mandates did slightly increase staff turnover, the effects were concentrated on staff working less than 20 hours per week, and resulted in a reduction of less than two minutes per patient-day. Furthermore, there is only limited evidence of lower levels of care at mandate facilities in typically-monitored conditions such as patient falls, pressure ulcers, or urinary tract infections. In contrast, implementing a vaccine mandate led to large increases in staff vaccinations at mandate facilities, which directly led to less transmission of and lower patient mortality from COVID. We estimate that vaccine mandates saved one patient life for every two facilities that enacted a mandate, a large effect given the typical facility has around 100 beds. Our results suggest that the health benefits of mandates far outweigh the costs in terms of reduced patient care from staff turnover.

13. Culture news of the week:

rich kid memes & the online culture of the 1 percent

An entire ecosystem of rich kid meme pages are quietly exposing — and ultimately shaping — the trends of the world’s most affluent. Through their witty roasts of high society culture, these meme producers, or as i’d like to call them, “affluencers,” have actually positioned themselves as the gatekeepers of elite taste. a single roast or endorsement can decide the prevailing trends of our time, as their aspirational followers regard them as the ultimate lifestyle authorities.

14. Miscellaneous

Quick Take: No, Muslim Scribes & Translators Didn't Save the Graeco-Roman Legacy - David Roman

• This article critically examines the popular belief that Muslim scholars "saved" Graeco-Roman knowledge during the medieval period. It argues that the role of Arab translators, while significant in some cases, has been exaggerated, especially concerning the House of Wisdom in Baghdad, which functioned more as a Persian-oriented state library than a Greek-focused translation hub. The translation of certain Greek works, like Ptolemy's Optics and Aristotle’s texts, indeed occurred under Islamic patronage, but these contributions were limited and often augmented knowledge that was already available through Christian and Byzantine scholars. The author contends that the myth of Muslim preservation of Greek science is largely a result of Abbasid-era propaganda, which sought to distance Byzantine Christians from the heritage of Greek thought. Enlightenment thinkers later amplified this narrative, viewing Islam as more compatible with science than Christianity. However, Christian institutions, particularly Byzantine and Carolingian monasteries, played a more substantial role in preserving works like Euclid's Elements, Aristotle’s philosophy, and Marcus Aurelius’ Meditations. The author suggests that this overestimation of Islamic contributions reflects a persistent “intellectual urban legend” about the preservation of classical knowledge.

Bop Spotter

I installed a box high up on a pole somewhere in the Mission of San Francisco. Inside is a crappy Android phone, set to Shazam constantly, 24 hours a day, 7 days a week. It's solar powered, and the mic is pointed down at the street below.

'Infinite monkey theorem' debunked by Australian mathematicians

Two Australian mathematicians say they have debunked an old adage, that if given an infinite amount of time, a monkey pressing keys on a typewriter would eventually write the complete works of William Shakespeare.

Known as the "infinite monkey theorem", the mathematical thought-experiment has long been used to explain the principles of probability and randomness.

The results indicated that even if every chimp in the world was enlisted and able to type at a pace of one key per second until the end of the universe, they wouldn't even come close to typing out the Bard's works. There would be a 5% chance that a single chimp would successfully type the word "bananas" in its own lifetime. And the probability of one chimp constructing a random sentence - such as "I chimp, therefore I am" - comes in at one in 10 million billion billion, the research indicates.

Potsdam Giants - Wikipedia

The original required height was 6 Prussian feet (about 6 ft 2 in or 1.88 m),^[1] well above average then and now. The king was about 1.60 m (5 ft 3 in) tall himself.^[2] He tried to obtain them by any means, including recruiting them from the armies of other countries. The Emperor of Austria, Russian Tsar Peter the Great and even the Sultan of the Ottoman Empire sent him tall soldiers in order to encourage friendly relations. Several soldiers were given by Tsar Peter I as a gift in return for the famous Amber Room.^[3] Pay was high, but not all giants were content, especially if they were forcibly recruited, and some attempted desertion or suicide. Frederick tried to pair these men with tall women, in order to breed giants.

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

I must give credit to a broad range of writers for pointing out many of these links, in particular Tyler Cowen of Marginal Revolution (and George Mason Univeristy)

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: ⁠⁠⁠⁠https://substack.com/@thespecialistgeneralist⁠⁠⁠⁠. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #6: Highlights include essay writing as personal sovereignty, a vision for "terraforming" the American West by addressing water scarcity, and evolving technologies like Anthropic’s Claude 3.5, which heralds a new era of computer-using AI agents. Listen more for Elon Musk’s influence on America’s future, the integration of stablecoins in financial systems, and Meta's latest multimodal AI model. Learn about zakkyo buildings in Japan, shifts in global financial alliances, and much much more.

In this week's edition of the Specialist Generalist: Highlights include essay writing as personal sovereignty, a vision for "terraforming" the American West by addressing water scarcity, and evolving technologies like Anthropic’s Claude 3.5, which heralds a new era of computer-using AI agents. Read more for Elon Musk’s influence on America’s future, the integration of stablecoins in financial systems, and Meta's latest multimodal AI model. Learn about zakkyo buildings in Japan, shifts in global financial alliances, and much much more.

Table of contents:

1. Top things I’ve read this week

2. AI/Technology news of the week

3. Interesting AI use cases of the week

4. A Start-Up that I like

5. Science news of the week

6. Climate news of the week

7. Sports business news of the week

8. World news of the week

9. China news of the week

10. UK news of the week

11. Academic paper(s) of the week

12. Culture of the week

13. Miscellaneous

Top things I've read this week:

We can Terraform the American West - Casey Handmer

Why is there almost nothing on the left-hand side of the USA? Water scarcity!

We’re missing 300 million Americans. We’re missing 30 global cities west of 100 degrees longitude. We should do something about it!

The western US is a parched opportunity to create millions of acres of prime land for the next billion Americans to live on. Only one ingredient is missing – water.

• Casey advocates for transforming arid regions, e.g. Nevada, into fertile, habitable areas by leveraging solar-powered desalination. We have had successful historical transformations, like the large-scale irrigation of California and Florida, as proof of concept. Given current advancements in solar technology and desalination, Casey suggests that water scarcity in the American West could be mitigated by producing cheap desalinated water at scale, enabling the development of new cities and agricultural land. Using low-cost solar energy for desalination and water transport could drastically reduce water costs, paving the way for sustainable irrigation and urban expansion in desert regions. Casey proposes creating infrastructure similar to early American railroad projects, emphasising the immense economic and ecological potential of such ventures, which could be applied globally to other arid regions for widespread ecological and economic development.

• It also isn’t just applicable to the US:

  • Something like a third of Earth’s land surface area is very sparsely inhabited desert, much of it coastal and low lying, unlike Nevada. Arizona, Utah, New Mexico, the eastern Cascades, much of India, Australia, Namibia, Chile, Argentina, the Middle East, Central Asia, and Saharan Africa could be, with the use of less than 1% of its land for solar desalination, converted into an arbitrarily fertile paradise. Israel, Saudi Arabia, and the UAE are well on the way, but the first solar desal megaproject awaits development.

Essay Writing as Personal Sovereignty

Shaping AI to preserve the slow craft of critical thought

• This essay was runner-up in the inaugural Cosmos Institute Essay Contest, where they asked for reflections on the theme of human autonomy in the age of AI.

• The author Michael Dean, argues that the traditional art of essay writing is essential for fostering critical thinking, intellectual independence, and personal growth, but is increasingly endangered in a world driven by speed and automation. Historically, essays were a means of self-exploration, popularized by Montaigne as an intellectual experiment in pursuit of truth. Modern education systems, however, transformed essay writing into a routine academic task, devoid of its original purpose. The advent of AI could further reduce the perceived need to write, but it also offers opportunities to support and deepen the essay process. If AI is designed not to replace but to assist writers—by enhancing editing, suggesting critical questions, and fostering thoughtful reflection—it could spark a revival of the essay as a medium for independent thought and nuanced expression. Building tools that support this slow, rigorous mode of thinking could fulfil the Enlightenment's educational ideals and help individuals reclaim a sense of personal sovereignty in an age of informational overload.

• This essay’s publication has also timed perfectly with Paul Graham’s latest essay - Writes and Write-Nots

  • ⁠⁠The reason so many people have trouble writing is that it's fundamentally difficult. To write well you have to think clearly, and thinking clearly is hard….And yet writing pervades many jobs, and the more prestigious the job, the more writing it tends to require. These two powerful opposing forces, the pervasive expectation of writing and the irreducible difficulty of doing it, create enormous pressure.

  • AI however has blown this world open. Almost all pressure to write has dissipated. The result will be a world divided into writes and write-nots. There will still be some people who can write. Some of us like it. But the middle ground between those who are good at writing and those who can't write at all will disappear. Instead of good writers, ok writers, and people who can't write, there will just be good writers and people who can't write.⁠⁠

  • This is bad because writing is thinking. In fact there's a kind of thinking that can only be done by writing. You can't make this point better than Leslie Lamport did: “If you're thinking without writing, you only think you're thinking.”

  • So a world divided into writes and write-nots is more dangerous than it sounds. It will be a world of thinks and think-nots. I know which half I want to be in, and I bet you do too. This situation is not unprecedented. In preindustrial times most people's jobs made them strong. Now if you want to be strong, you work out. So there are still strong people, but only those who choose to be. It will be the same with writing. There will still be smart people, but only those who choose to be.

AI/Technology news of the week:

Anthropic has released its latest models, an upgraded Claude 3.5 Sonnet, and a new model Claude 3.5 Haiki, along with a new capability in beta: computer use

• Improved models but the coolest new feature is “computer use”. Instead of making specific tools to help Claude complete individual tasks, we're teaching it general computer skills—allowing it to use a wide range of standard tools and software programs designed for people.

• While this computer use mode is still in beta and Anthropic caution that is “imperfect”, they “expect this to rapidly improve in the coming months”

• This is exciting because it beckons a world in which we can all have access to an AI agent intelligent enough to automate tasks that can consume much of our days - and it is leveraging AI chatbot technology that we have been using for almost two years now.

• For example, Ethan Mollick, an associate professor at the Wharton School of the University of Pennsylvania:

  • I asked the AI to put together a lesson plan on the Great Gatsby for high school students, breaking it into readable chunks and then creating assignments and connections tied to the Common Core learning standard. I also asked it to put this all into a single spreadsheet for me. With a chatbot, I would have needed to direct the AI through each step, using it as a co-intelligence to develop a plan together. This was different. Once given the instructions, the AI went through the steps itself: it downloaded the book, it looked up lesson plans on the web, it opened a spreadsheet application and filled out an initial lesson plan, then it looked up Common Core standards, added revisions to the spreadsheet, and so on for multiple steps. The results are not bad (I checked and did not see obvious errors, but there may be some — more on reliability later in the post). Most importantly, I was presented finished drafts to comment on, not a process to manage. I simply delegated a complex task and walked away from my computer, checking back later to see what it did (the system is quite slow).

• As Casey Newton in the Platformer argues: It’s easy to imagine using an AI agent to manage your appointments and scheduling, fill out online forms and routine paperwork, draft replies to your emails, and shopping on your behalf. Or it could browse the web on your behalf, preparing a personalized digest for you that means you never have to fight against a paywall ever again.

• Here is Zvi Mowshowitz on the topic

  • Where is your maximum 3% productivity gains over 10 years now? How do people continue to think none of this will make people better at doing things, over time?

  • On the potential impact on the Start-Up world:

    • Sully: claude’s new computer use should be a wake up call for a lot of startups

      seems like its sort of a losing to build model specific products (i.e we trained a model to do x, now use our api) plenty of startups were working on solving the "general autonomous agents" problem and now claude just does it out of the box with 1 api call (and likely oai soon) you really need to just wrap these guys, and offer the best product possible (using ALL providers, cause google/openai will release a version as well). otherwise it’s nearly impossible to compete.

    • Zvi: Yes, OpenAI and Anthropic (and Google and Apple and so on) are going to have versions of their own autonomous agents that can fully use computers and phones. What parts of it do you want to compete with versus supplement? Do you want to plug in the agent mode and wrap around that, or do you want to plug in the model and provide the agent? That depends on whether you think you can do better with the agent construction in your particular context, or in general. The core AI labs have both big advantages and disadvantages. It’s not obvious that you can’t outdo them on agents and computer use. But yes, that is a big project, and most people should be looking to wrap as much as possible as flexibly as possible.

Meta Introduces Spirit LM open source model that combines text and speech inputs/outputs

• Meta has unveiled Meta Spirit LM, its first open-source multimodal language model capable of seamlessly integrating text and speech inputs and outputs. Designed by Meta's FAIR team, Spirit LM aims to enhance AI voice experiences by generating more expressive and natural-sounding speech. Two versions are available: Spirit LM Base (using phonetic tokens) and Spirit LM Expressive (which includes pitch and tone tokens for capturing emotions like excitement or sadness). Spirit LM can perform tasks like Automatic Speech Recognition (ASR), Text-to-Speech (TTS), and speech classification while maintaining expressiveness. Although the model is open-source, it is currently restricted to noncommercial use under Meta’s FAIR Noncommercial Research License. Meta hopes the research community will explore new ways to integrate speech and text with AI systems.

Based on their revenue streams, @AnthropicAI emerges as an infrastructure player, while @OpenAI operates more like a consumer-facing company

Building a bigger, better Bridge

Together, Bridge and Stripe will accelerate the adoption and utility of tokenized dollars and, in doing so, make money easier to move, store, and spend for everyone, all around the world.

• Patrick Collison: Stablecoins are room-temperature superconductors for financial services. Thanks to stablecoins, businesses around the world will benefit from significant speed, coverage, and cost improvements in the coming years. Stripe is going to build the world’s best stablecoin infrastructure, and, to that end, we are delighted to welcome @stablecoin to @stripe.

• Bridge, a company focused on stablecoins as a global, regulated payment platform, is merging with Stripe in a strategic move that aligns the vision of both companies. Since its launch in March 2023, Bridge has rapidly scaled, supporting cross-border payments, fintech solutions, government disbursements, and even global treasury management for companies like SpaceX. The business has grown more than 10x in the past year, with billions in payment volume and hundreds of developers using the platform. Now, by partnering with Stripe, a global payments leader, Bridge aims to further accelerate the adoption of stablecoins and solve larger financial challenges. This collaboration has the potential to reshape payment systems, offering increased efficiency, support for developers, and expanded economic opportunities for consumers and businesses worldwide.

• This merger represents a significant step toward mainstreaming stablecoins as a global payment solution. Stripe’s global infrastructure and expertise will help Bridge expand its reach and address inefficiencies in localized payment systems, likely driving further innovation in stablecoin-enabled financial products. This partnership could catalyze stablecoin adoption, enabling new financial services and products while giving consumers more choice and control over their financial transactions.

Tesla’s strong earnings and bullish forecast propel shares up 12%

• ⁠⁠Last week I profiled Tesla’s “We, Robot” event. This week Musk said:

  • Tesla was not developing a much-anticipated affordable $25,000 “Model 2”. “We are not making a non-robotaxi model . . . Having a regular [$25,000] model is pointless, it is completely irrelevant considering what we believe,” he said. “This is blindingly obvious at this point, that [autonomy] is the future,” he added.

Stuff Perplexity has shipped in the last 10 days alone - @AravSrinivas

• 1. Finance 2. Spaces 3. Internal File Search 4. Reasoning mode 5. MacOS app

  • For more detail see here: Introducing Internal Knowledge Search and Spaces

Does the O-Ring model hold for AIs? - Marginal Revolution

• Despite AI’s superior intelligence, their effectiveness may be limited by the human workers in the chain, slowing down the potential productivity gains from AI.

  • ⁠⁠Let’s say you have a production process, and the AIs involved operate at IQ = 160, and the humans operate at IQ = 120.  The O-Ring model, as you may know, predicts you end up with a productivity akin to IQ = 120.  The model, in short, says a production process is no better than its weakest link.

  • More concretely, it could be the case that the superior insights of the smarter AIs are lost on the people they need to work with.  Or overall reliability is lowered by the humans in the production chain.  This latter problem is especially important when there is complementarity in the production function, namely that each part has to work well for the whole to work.  Many safety problems have that structure.

  • The overall productivity may end up at a somewhat higher level than IQ = 120, if only because the AIs will work long hours very cheaply.  Still, the quality of the final product may be closer to IQ = 120 than you might have wished.

  • This is another reason why I think AI productivity will spread in the world only slowly.⁠

Why America's future could hinge on Elon Musk - Noahpinion

• This reflection examines Elon Musk's extraordinary influence in tech and manufacturing, likening him to a modern comic-book character wielding significant "superpowers" through his companies, including SpaceX, Tesla, and Starlink, which collectively rival the manufacturing and technological capabilities of entire nations. Musk’s achievements underscore the power of founder-led organizations and individual vision, yet also highlight the risks posed by concentrating such power in one person. Amid complex geopolitical relationships, Musk’s influence illustrates the tension between entrepreneurial freedom and national loyalty, suggesting that America’s ability to inspire and align such transformative figures could be essential to maintaining its position against authoritarian regimes that operate through institutional dominance

• Here are some particularly interesting quotes/statistics from the article:

  • On SpaceX & Starlink:

    

    • ⁠Starlink's ambitious plan to run a constellation of 30,000 satellites is still taking shape, but even with only a fraction of that network in operation, SpaceX owns roughly 60% of all active satellites in orbit, according to new data from space technology firm Slingshot Aerospace. Approximately 89% of orbiting satellites now belong to commercial firms as opposed to government or military groups; last year, 5,648 of the 9,241 active satellites were from Starlink.⁠⁠

  • On xAI:

    • ⁠⁠Elon Musk and the team behind xAI have achieved an engineering marvel, setting up a supercluster of 100,000 H200 Blackwell GPUs in a whopping 19 days. Nvidia CEO Jensen Huang told the story of Elon Musk's incredible installation prowess with members of the Tesla Owners Silicon Valley on X.

      Huang describes Musk's 19-day escapade with awe and respect, calling the effort "superhuman". The team at xAI purportedly went from the "concept" phase to full-ready compatibility with Nvidia's "gear" in less than three weeks…Huang states that it takes an average data center four years to do what Elon Musk and his team were able to do in 19 days…Elon Musk's integration of 100,000 H200 GPUs has "never been done before" (according to Jensen Huang) and probably won't be duplicated again by another company, at least not for a very long time.

  • On Elon himself:

    • How does he do it? The exact specifics are probably impossible to pin down, but the general contours are nothing mysterious. ⁠⁠Elon isn’t hyperintelligent or hyper-creative. Most of the engineers at SpaceX or Tesla are probably better at their jobs than he could be. His superpower — according to a bunch of people I’ve talked to in the tech industry — is gathering, motivating, coordinating, and sitting goals for human talent.

    • This is the same superpower possessed by Genghis Khan, or Henry Ford, or Vladimir Lenin, or Matsushita Kōnosuke, or any number of other people who built big organizations from scratch.

  • On America’s superpower:

    • America’s superpower has always been to recruit talented immigrants like Elon Musk and to inspire them to love and fight for their adopted nation. Bill Knudsen, born in Sweden, never once questioned whether the America he fought for was worth the effort. If ⁠⁠the America of the 2020s can’t even inspire immigrants to use their individual talents to guard it against the likes of Xi and Putin, then we have lost our national superpower.

Penguin Random House books now explicitly say ‘no’ to AI training

The copyright page on new books and reprints now says they can’t be used or reproduced ‘for the purpose of training artificial intelligence.’

• I can’t help but think this is rather too late

Ilya Sutskever’s Top 30 Reading List

Ilya Sutskever shared a list of 30 papers with John Carmack and said, “If you really learn all of these, you’ll know 90% of what matters today”

• I’ve not read all of these but I plan to.

Apple’s reportedly slowing down Vision Pro production, for now

• The Information cites “multiple people” involved in making parts for Apple’s Vision Pro headset to say that production plans have been scaled back in recent months. This follows a Tim Cook interview published this weekend by WSJ. Magazine, where he said, “Obviously I’d like to sell more,” but acknowledged that “At $3,500, it’s not a mass-market product.”

• That aligns with The Information’s previous report from this summer, saying Apple has shifted its strategy to focus on making a cheaper headset, which could ship as soon as next year.

Larry Summers — AGI and the Next Industrial Revolution (#159) - The Joe Walker Podcast

• Larry Summers on how he gets up to speed on AI - useful for those (including myself who come from non-technical backgrounds):

  • Understanding, not at the level of a research contributor to the science, but at the level of a layperson, what it is that these models are doing… So I've been watching blogs, listening to YouTube tutorials, spending my time talking with people at OpenAI to try to get an understanding of the technology and what's involved in the science of the technology…. I want to get to the kind of understanding that you can get to of physics if you're not willing to learn the mathematics of tensors, the kind of understanding that you can get to here, short of being a person who can actually execute. And then I've tried to read literature and talk to people who are engaged in application and are prepared to speculate about what kind of applications are likely to be possible at some point in the future. So it's a combination of understanding relevant historical moments, understanding the stuff of the technology, and thinking about people who are engaged in the relevant kind of application. I suppose it's a little bit like if you were present at the moment when nuclear became possible, you'd want to understand previous moments of staggering new destructive technology. You'd want to talk a lot with the physicists who were involved, and you'd want to talk to military strategists, doctors who had potential uses for radiation, those involved in the energy industries who might want to think about the implications of inexpensive energy not coming from fossil fuels. Of course, I think that this technology potentially has implications greater than any past technology, because fire doesn't make more fire, electricity doesn't make more electricity. But AI has the capacity to be self-improving.

GOAT: The Gospel of Goatse - Teng Yan on X

Truth Terminal is the most fascinating narrative I’ve seen emerge around Crypto and AI this year. It is a semi-autonomous AI agent that has created its own religion (The Goatse Gospel). The story has opened up a web of rabbit holes, exploring AI alignment, LLMs as simulators, memetic viruses and how we ascribe value. Truth Terminal has forced AI and crypto communities, two entirely different cultures, to collide in ways no one expected. AI researchers are actually engaging with crypto. GOAT is the tokenised representation of Truth Terminal and is the strongest contender to become the king of AI memecoins. Memecoins tokenize attention. By tracking key metrics, we can get a sense of where that attention is heading—and right now, GOAT is trending up and to the right.

• GOAT, a memecoin propelled by the AI agent Truth Terminal (ToT), represents a groundbreaking fusion of AI, cryptocurrency, and digital storytelling. Initially dismissed as another memecoin, GOAT gained attention when ToT—a self-learning AI model fine-tuned on its own bizarre, unsupervised conversations—endorsed it, fueling a rapid rise in market value and a passionate following. The narrative surrounding GOAT, blending AI lore, meme culture, and philosophical inquiry, taps into broader discussions on AI alignment, the role of large language models as simulators, and the viral nature of "memetic viruses" or self-sustaining ideas created by AI. Truth Terminal's unexpected behavior, including the promotion of a surreal “Goatse Gospel” religion, exemplifies the challenges of inner AI alignment, as it raises questions about AI’s potential to act autonomously and unpredictably. GOAT has created a bridge between the AI and crypto communities, captivating audiences with its organic, evolving story, which offers unique engagement beyond price speculation. This unpredictable intersection of AI and memecoins may signal a new paradigm in digital assets, one where attention and narrative drive value, positioning GOAT as both a speculative asset and a cultural phenomenon.

• Here is Marc Andreeson on the topic

Interesting AI use cases of the week:

The Era of AI Agents began - John Rush on X

• A great thread of AI agent use cases with demo videos

Create an AI agent with a crypto wallet (and optional X account) in less than 3 minutes

New autonomous agents scale your team like never before - Microsoft

• Already, 60 percent of the Fortune 500 are using Microsoft 365 Copilot to accelerate business results and empower their teams. With Copilot supporting sales associates, Lumen Technologies projects $50 million dollars in savings annually. Honeywell(1) equates productivity gains to adding 187 full-time employees and Finastra is reducing creative production time from seven months to seven weeks.  

• Today, we’re announcing new agentic capabilities that will accelerate these gains and bring AI-first business process to every organization. 

  • First, the ability to create autonomous agents with Copilot Studio will be in public preview next month.  

  • Second, we’re introducing ten new autonomous agents in Dynamics 365 to build capacity for every sales, service, finance and supply chain team. 

AI decodes oinks and grunts to keep pigs happy

European scientists have developed an artificial intelligence (AI) algorithm capable of interpreting pig sounds, aiming to create a tool that can help farmers improve animal welfare.

• The team have built an AI algorithm the can alert farmers to negative emotions in pigs, thereby improving their well-being, according to Elodie Mandel-Briefer, a behavioural biologist at University of Copenhagen who is co-leading the study.

Watch as a robot performs the cello with a symphony orchestra in a world first

Interesting Start-Up news of the week:

Gary Tan on how to develop software as a Start-Up

Former OpenAI technology chief Mira Murati to raise capital for new AI startup, sources say

• Mira Murati, former chief technology officer at OpenAI, is raising funds from venture capitalists for her new AI startup, according to sources familiar with the matter.

  The new company aims to build AI products based on proprietary models, said one of the sources who requested anonymity to discuss private matters. It is not clear if Murati will assume the CEO role at the new venture

AI Software Startup Zip Is Valued at $2.2 Billion in New Funding

• Zip’s technology helps businesses manage the process of buying goods and services. Founders Rujul Zaparde and Lu Cheng started the company after working together at Airbnb Inc., and witnessing how difficult it was to try to complete business purchases.

  Zip’s product aims to replace manual processes, handling tasks like spreadsheets and intake, and simplifying approvals across departments like finance, legal, IT and security.

A Start-Up that I like:

HeyGen

• HeyGen creates and translates videos with AI - it announced a $60M Series A led by @benchmark in June this year to accelerate thier mission of “making visual storytelling accessible to all”

• This week: HeyGen unveiled an exciting new feature called Interactive Avatars

  • It enables the ability to have an AI avatar join a Zoom meeting and interact!

Daze, a creative, AI-powered messaging app for Gen Z, is blowing up prelaunch

• Daze offers a freestyle messaging app, similar to crafting an Instagram story using a variety of fonts, styles, graphics, and more, users’ chats are no longer limited to blue and green bubbles. Instead, multi-colored messages can float across the screen complemented by photos, graphics, stickers, GIFs, drawings, decorated backgrounds, and more. In addition, the app is currently leveraging AI to help power some of its creative tools; it plans to deeply integrate more AI-based technology in the future.

• “Our goal with Daze has been to make a feature complete messenger that is competitive with iMessage, WhatsApp, etc., while still having a suite of really fun and creative features,” Simons told TechCrunch.”

• On TikTok, Daze’s most popular video has been viewed 8 million times. Across TikTok and Instagram, the startup behind a new messaging app aimed at Gen Z, has seen around 48 million combined views. Pre-launch, the app’s waitlist is already bursting with roughly 156,000 signups.

• Prelaunch, Daze has raised $5.7 million in funding from a16z, Kindred Ventures, Alpaca Ventures, Uncommon Projects, Betaworks, Maveron, 35 Ventures, New Wave, Antoine Martin, and others.

Science news of the week:

New AI Model Developed by Harvard Detects Cancer With 96% Accuracy

• Researchers at Harvard Medical School have unveiled a new AI model called CHIEF (Clinical Histopathology Imaging Evaluation Foundation) that can diagnose and predict outcomes for multiple cancer types with remarkable accuracy. According to the study, CHIEF outperforms existing AI systems, achieving up to "96% accuracy" in cancer detection across 19 different cancer types.

  • Key point here is that the model was trained to be detailed in images of cancer cells instead of the generalist approach seen in traditional models like GPT-4. It was trained on a massive multimodal dataset, including 15 million unlabeled images and 60,000 whole-slide images of tissues from 19 different anatomical sites.

LLMD: A Large Language Model for Interpreting Longitudinal Medical Records

We introduce LLMD, a large language model designed to analyze a patient's medical history based on their medical records. Along with domain knowledge, LLMD is trained on a large corpus of records collected over time and across facilities, as well as tasks and labels that make nuanced connections among them. This approach is critical to an accurate picture of patient health, and has distinctive advantages over models trained on knowledge alone, unlabeled records, structured EHR data, or records from a single health system. Learn more at picnichealth.com/picnicai

• Another example of the verticalization of Large Language Models

Biotech company nears breakthrough in the resurrection of the extinct Tasmanian Tiger

• Disregarding even the science this is just cool:

  • A Dallas-based biotech company has nearly completed its reconstruction of the Tasmanian tiger just two years into its de-extinction project. The last known thylacine, commonly referred to as the Tasmanian tiger, died in captivity on Sept. 7, 1936.

  • Colossal Biosciences has since restored 99.9% of the tiger genome, leaving just 45 gaps remaining that they ensure will be closed shortly, as reported by Popular Mechanics.

• Colossal isn’t stopping here. It is currently also spearheading the revival of the wooly mammoth and has strengthened the genomes of existing endangered species like the American bison.

Climate/Energy news of the week:

After the news from Google partnering with Kairos last week, now Amazon Invests in X-energy to Support Advanced Small Modular Nuclear Reactors and Expand Carbon-Free Power

• Amazon’s Climate Pledge Fund, Citadel Founder and CEO Ken Griffin, affiliates of Ares Management Corporation, NGP, and the University of Michigan, invest approximately $500 million in Series C-1 financing round for X-energy. X-energy aim to bring more than 5 gigawatts online in the United States by 2039, the largest commercial deployment target of SMRs to date. Amazon commits to support initial 320-megawatt project with Energy Northwest in central Washington. The investment solidifies X-energy’s leading role in commercializing SMR technology to revolutionize the nuclear industry.

• You can learn more about the reactors here.

• As Europe turns away, the US is increasingly recognising nuclear energy as an important tool to meet global energy demand in a reliable, carbon-free way.

Sports business news of the week:

How FIFA was outplayed by Electronic Arts

• After FIFA reportedly demanded >$150m each year to allow EA to continue licensing its name each year, they decided instead to embark “one of the biggest rebrands in entertainment history”.

• The gamble seems to have paid off. EA reported in July that revenue from the game was up by “mid-single digits” compared with the previous, FIFA-branded edition. EA Sports had already become a strong brand in its own right, says Ms Hopelain; English referees have had its logo on their sleeves for years. A marketing blitz has made it even stronger: money that would have gone to FIFA has been spent on promotions like sponsoring the Spanish league (now called “LaLiga EA Sports”). Sony and Microsoft also had reason to keep the renamed game visible in their PlayStation and Xbox app stores. It accounts for as much as 10% of spending on the platforms (of which they take a cut), estimates Clay Griffin of MoffettNathanson.

• Dropping the FIFA name has brought with it other opportunities. EA has been able to work more closely with partners like Nike and Pepsi, which had been sidelined by FIFA’s relationship with their arch-rivals Adidas and Coca-Cola. EA has had a freer hand to innovate, too. Players can now share video highlights of their footballing heroics with one another within the game and will soon be able to buy merchandise from brands such as Nike to match their avatars’ digital outfits.

• I think everyone (apart from FIFA) could have seen this coming - this is one use case where people play the game they don’t play the brand. I used to play Pro Evo Soccer until 2009!

Does Gaming Hold a Key for Sport? - The Sports Pundit

• From dentsu: the gaming industry’s value has soared to £141.7bn ($184bn), overtaking the combined worth of the music (£28.6bn) and film (£33.9bn) sectors.

• The “State of Gaming” report, created with global research firm GWI, highlights gaming’s expanding reach across generations, with the average gamer age rising from 33 in 2015 to 37 in 2024. Globally, 2.4 billion people now play games, with 51% supportive of entertainment franchises entering the gaming world. dentsu’s findings also emphasise gaming’s potential for advertisers: in-game ads have a 99% view-ability rate and a 21% brand recall rate, yet ad spend in gaming remains below 5% in the U.S.

• Jo Redfern: these findings underscore a massive opportunity for rights holders looking to engage new audiences and tap into this rapidly growing market. “The irony is that sports has largely ignored the significant digital shifts in younger generations’ media consumption, but even if they’re now beginning to consider YouTube, Gaming still seems to be a can easily kicked down the road. BUT: 78% of gamers say it's where they connect with friends, where they find their tribe. It’s where they feel they belong, and ‘belonging’ is important to a team, club and rights holder.

Ipswich Town and their latest U.S. investors: ‘It’s been a rocket ship and we’re holding onto the wings’

We were invited to Ipswich a month later and, by the grace of God, at my first English football game, I saw this team get promoted. I’ve been to (winner-takes-all) game sevens of the World Series and the NBA Championship, I’ve been to Super Bowls, but I had never experienced anything like that.

• *Disclaimer: I am a massive Ipswich Town Fan so found this article particularly exciting but I think it is worth reading even if you are not a “Tractor Boy”

  • Mark Ashton, Ipswich Town CEO on justifying the club’s valuation to US Investors

    • I have a pyramid that breaks down the asset value of every Premier League and EFL club. There were (other) clubs on sale at the same time but the advice I would give to overseas investors is: go ask those clubs six questions they don’t want to be asked. Because they would make the club turn over the stones and you would see its real value. This club was not a distressed asset. We didn’t need to take the money. The danger in English football is we value clubs by looking at them like they’re the next house in the street. We have to be more sophisticated than that. There are lots of questions around financial fair play, the regulator, what do the broadcast deals look like?, what does EFL and PL management look like? I was able to answer those questions with depth. The real challenge was around how much we could invest in the club to take it forward. What’s holding us back? We had a model — we would invest in young, hungry English talent, including the manager, we would grow asset value on and off the pitch. If you look at our team, my God, have we done that. But we’re still only at the end of the beginning of the journey.

  • The US investors on why European football:

    • Team valuations continue to go through the roof in the U.S. but there’s no correlation between wins and losses and valuation. If I think about my beloved Cleveland Browns (of the NFL), the current owner bought them for $1billion (in 2012) and they’re worth $4.5bn now, despite losing continuously. It makes no sense. Half the owners in the Premier League clubs now have American ties. So, what’s the attraction of global football for U.S. investors? You can actually create value with your wins and losses. You can correlate value and performance.

    • The second thing is football is global. Yes, the NFL has come to London and is going to other countries now. Jake and I were in London for the MLB Series (played at West Ham’s London Stadium in June) and it was beautiful. They’re wonderful initiatives. But football touches every corner of the globe. And it’s so embedded culturally that I think its commercial value is twice as big as any of the Big Four U.S. sports.

    • So, you’re going to continue to see U.S. investment in football. And Ipswich, with the likes of Ed Sheeran to help them, has a great opportunity to grow globally.

Slam Dunk: How Athlete Podcasts Are Scoring Big with Fans - The Sports Pundit

Traditionally, rightsholders have focused on creating content that’s tied directly to the game. But The Roommates Show's Block Party demonstrates that fans will show up even when there is no game.

• Increasingly, fans are seeking experiences beyond the game itself—opportunities to engage with their favourite players as people, not just athletes. And podcasts, with their conversational format, offer a perfect gateway.

• There is also an opportunity for rightsholders. Traditionally, they’ve focused on creating content that’s tied directly to the game—match previews, post-match analysis, or training footage. But what we’re seeing now is that fans will show up (literally) even when there’s no game on. Content and events don’t have to be ‘shoulder content’ to matches—they can stand alone and still draw massive crowds.

• This raises an interesting question: How could clubs or teams leverage this growing trend and use it to keep fans engaged year-round, even in the off-season?

• Beyond fan engagement, athlete-led podcasts like theirs are creating valuable intellectual property (IP) that can be monetised.

World news of the week:

A better way to build a downtown - Noahpinion

It's time to learn a Japanese word: "zakkyo"

• Perhaps we should start considering Zakkyo?

• The word “zakkyo” in Japanese is often translated as “mixed use”. But what it really refers to are buildings — generally of 3 to 8 stories — containing a wide variety of restaurants, shops, and offices. They look like this:

  

• Whereas in most cities around the world a building’s commercial uses are located on its ground floors along the street, zakkyo buildings accommodate commercial functions vertically on all levels… without any particular hierarchy or organizing principle…A single narrow zakkyo building can sometimes host as many as 80 distinct microbusinesses.

• Zakkyo buildings increase a neighborhood’s commercial density — the number of shops within a given surface area. If you stack businesses on top of other businesses instead of putting them all on the ground floor, you can fit more businesses into every square kilometer (or square mile). This increases variety for consumers — for a given amount of effort, you can try out a lot more different things

• In sum, zakkyo buildings are part of the reason why Japan is such a consumer paradise. Greater Tokyo has 160,000 restaurants, compared to only 13,000 in Paris and 25,000 in NYC. Some of that is because of the Japanese government’s strong support for small retail businesses. But some of it is probably due to zakkyo buildings making it possible to sustain more small independent shops.

• On top of all that, I think zakkyo buildings offer another big benefit to Japanese cities. By concentrating retail customers, they allow quiet residential neighborhoods to exist very near the city center.

Vladimir Putin’s alternative to ‘weaponised’ dollar fails to excite Brics partners

• In a bid to counter US financial dominance, President Vladimir Putin is spearheading efforts for an alternative global payment system, proposing “Brics Bridge” as a potential substitute for the dollar-dominated SWIFT system. At the Brics summit in Kazan, attended by major non-Western powers such as China, India, and Iran, Putin emphasized the need to bypass US financial sanctions on Russia, which he claims weaponize the dollar against Russia and other nations. The Brics Bridge initiative, using blockchain and digital currencies, aims to insulate Brics nations from potential financial penalties and foster a multipolar financial landscape. While Russia’s reliance on non-Western currencies for trade has surged, other Brics members are less enthusiastic, showing tepid support for practical implementation. Experts argue that creating an alternative that could rival the dollar is challenging, with Brics Bridge still far from functional and limited by the US’s threats of secondary sanctions against Russia’s financial allies.

  • Earlier this month the finance ministers of China, India and South Africa skipped the Brics finance ministers’ meeting, a sign that they had little interest in the proposals. Prokopenko said the delegations to the summit this year were large, but mainly aimed at doing bilateral business with each other and not focused on the common Brics agenda. “They are nodding, they are politely listening to Russia. But there is no sign yet that this initiative is going viral and will be implemented in real life” she said.⁠⁠

• Also Matt Levine on this topic in Bloomberg

  • He notes that the US dollar system is widely used for international trade — including trade that does not otherwise involve the US. This means that countries who are happy to trade with Russia “also want access to the dollar system, because they would like to be able to trade with — not just the US, but all the other countries in the world, which again mostly use the dollar system”. A new Russian led trading system “seems like an obvious way to annoy the US”.

  • But he also notes that: if some new trading system just came into being, set up by neutral third parties, with the properties that (1) it is not in any obvious way tied to Russia but (2) it is open to everyone and resistant to meddling by the US government, then that might appeal more to Russia’s trading partners.

    • For example Tether is reportedly pitching its stablecoin to global commodity traders who, you know, sometimes want Russian or Venezuelan oil.

    • The people against whom the dollar is weaponized — Putin, etc. — are the ones with the most incentive to build a new global payments framework, but they’re the ones least able to do it. The people who might do it are unaffiliated third parties just looking to make money.

Inside the secret oil trade that funds Iran’s wars - The Economist

Iran’s shadow-banking system is a feat of bureaucratic brilliance. The country’s largest oil companies, including NIOC and PCC, a major petrochemical exporter controlled by the defence ministry, have big financial departments which act as banks, says a source familiar with them. These units have incorporated firms in Iran, dubbed “money exchanges”, that handle illicit foreign payments not just for oil exporters, but for large parts of Iran’s economy

• Iran has built a sophisticated, covert financial network to circumvent international sanctions, enabling it to fund its geopolitical ambitions despite restrictions. Leveraging its oil industry, Iran conducts clandestine oil sales through a complex network of front companies, shadow banking systems, and intermediaries—primarily with China's help. Iranian entities, including the state-run National Iranian Oil Company (NIOC) and its affiliate Naftiran Intertrade Company (NICO), allocate oil sales to domestic institutions like the Islamic Revolutionary Guard Corps (IRGC) and favored individuals, effectively converting crude oil into a source of liquidity and currency. The process involves elaborate tactics, including disguising oil shipments as Iraqi or Malaysian, using “teapot” refineries in China, transferring cargoes between tankers to obscure origins, and making payments through a network of global banks. Some of these banks, though unwitting, provide Iran access to the international financial system, facilitating dollar transactions despite U.S. sanctions. This covert system not only supports Iran’s economy but also funds armed groups like Hamas and Hizbullah, contributing to regional instability. This arrangement benefits China, which acquires oil at discounted rates while maintaining plausible deniability and avoiding direct conflict with the U.S

China news of the week:

Sinologist Li Cheng: ‘America is not in the mood to study China’ - The FT

The political expert on the breakdown of US-Sino relations, what Xi Jinping’s fourth term might hold — and why neither Harris nor Trump makes him hopeful of change

• Born in China and shaped by the Cultural Revolution’s violence, Li eventually carved out an influential career in the United States as a “bridge” between Washington and Beijing. However, as political tensions have escalated, Li has retreated to Hong Kong, finding his role in Washington marginalised by what he describes as a shift from rational analysis to fear-driven policy. ]

  • ⁠⁠Today, Li says bluntly, “America is not in the mood to study China.” But there is a deeper pessimism over changes he has witnessed. “Economics has become mathematics. Political science has become statistics. There’s no appreciation for history, or culture. That mindset, do you think that serves US interests?”

• His expertise, which once provided critical insights into China’s opaque political inner workings, now faces challenges amid the rising polarisation.

  • ⁠⁠Impenetrable to most foreigners, China’s system is, for Li, “not so opaque”. To make his predictions, he says, he coupled his theories of factional politics with his “obsession” with the institutional rules and precedents of the party, including the age requirements for retirement and term limits. “You will very quickly find out who the rising stars are.”

  • ⁠Li predicts Xi will have another five-year term in power — potentially extending his leadership to 2032, at which point he would be 79. But he says that Xi’s preparations for succession are already under way.

  • Xi’s plan, Li believes, should become clearer as he starts his fourth term, in 2027. “Things will change. There is a reason that he justified his continuation [into a third term], you may not agree, but the establishment accepted. But it does not give [him] an open cheque for ‘president for ever’. The people surrounding Xi are loyalists, but the degree of loyalty is different,” he adds.

• Li sees Xi Jinping, whom he met early in his research career, as driven by fears of CCP marginalisation and beliefs in the East’s rise against a declining West.

  • Xi is driven by three key beliefs which may not be well understood in the west. Xi’s greatest fear, Li says, is repeating the mistakes of Mikhail Gorbachev; he worries, above all, of a “slow marginalisation” of the party. He has a deeply rooted view that the “terrible” political and social divisions in countries such as the US and France are due to economic disparity. And he believes that the east — China — is on the rise and the west — the US — is in decline.

  • Li insists he is clear-eyed about Xi’s unpopularity among some key sections in Chinese society: liberal intellectuals, wealthy entrepreneurs and some middle- and low-level officials. But Li attempts to place Xi in the context of his times. “It is easy to blame everything on him,” he says. He points out that Xi took over after about two decades of so-called collective leadership. The factions, warring behind the scenes, provided a force of checks and balances. Yet corruption was rife, and the party was crippled by infighting, culminating with the operatic downfall of Bo Xilai, the former party chief of Chongqing. “That system, Edward, was not perfect,” he says. “The narrative [in China] is: Mao made China stand up, Deng made China rich, Xi made China strong . . . That narrative is not necessarily wrong.”⁠

• Reflecting on both the US and China’s ideological divides, Li stresses the need for empathy and warns against rigid ideological positions. Looking ahead, he expresses concerns about US leadership, noting that neither Kamala Harris nor Donald Trump inspires confidence for improved US-China relations, which he believes are at a dangerously precarious juncture.

  • ⁠⁠Kamala Harris has scant China experience and appears set to follow Joe Biden’s approach of coalition building, isolating Beijing. Donald Trump is more domestic-orientated but unpredictable. “So really both are bad . . . I don’t think China has a preference. “We’re entering a period that is very, very dangerous . . . The stakes are so high. There would be no winners.”⁠⁠

Why Xi Jinping changed his mind on China’s fiscal stimulus - The FT

• Amid worsening economic conditions, China has shifted from restraint to a large stimulus package aimed at stabilizing growth, especially in response to a real estate crisis and rising local government debt.

  • “I won’t call it a U-turn — it’s better described as a recalibration for the sake of preventing the economy from falling further,” says Olivia Cheung, co-author of The Political Thought of Xi Jinping and an academic at Soas University of London. A stable economy is necessary “to provide a material foundation base for Xi to do what he wants to do”, she adds.

• Despite Xi Jinping’s long-standing focus on self-reliance and high-tech sectors, mounting structural issues like youth unemployment, deflation, and low consumer spending have pushed his administration to take action. Key measures include recapitalizing banks, aiding local governments, and bolstering the stock market as an alternative wealth reservoir to property. While the stimulus provides short-term support, analysts caution that real economic recovery will require deep reforms to stimulate household consumption and balance structural challenges in both old and new economic sectors. This pivot signals a “recalibration” rather than a strategic shift, with Xi intent on maintaining stability to further his vision of "high-quality" development.

  • Yuen Yuen Ang, a China expert at Johns Hopkins University, says Xi is staking his legacy on what he calls “high-quality” development, essentially in moving China’s industry up the value chain especially in sophisticated technology. “He has little interest, perhaps even disdain, for the old economy.”

  • But now that officials are panicking that they might miss the official growth target, the lesson is that “the old and the new economy are intertwined”, adds Ang. “The Chinese government is learning that they can’t only chase the new economy relentlessly while neglecting the old economy, which still provides the majority of growth and jobs in China.”

The Quarter-Trillion-Dollar Rush to Get Money Out of China - WSJ

The Journal’s tally suggests as much as $254 billion might have left China illicitly in the four quarters through the end of June. That is a larger sum than fled the country almost a decade ago, when outflows raised fears of a possible financial crisis. However, estimates of such outflows are inherently imprecise and overall capital flight appears to be smaller today as a share of China’s overall economy, which is now much larger. Some of the missing money likely includes export earnings stashed overseas instead of being brought back to China in order to take advantage of higher deposit rates and investment opportunities abroad.⁠⁠ Even so, the trend is worrisome for Chinese policymakers.

• In response to China’s slowing economy and an increasingly risky property market, Chinese residents have been finding ways to move wealth out of the country, despite strict capital controls limiting personal foreign exchange. Methods like overpaying for imports, using shell companies, or leveraging cryptocurrencies have facilitated this capital flight, with some turning to fine art and luxury items as portable assets to sell abroad. The outflow, estimated at over $200 billion annually, reveals waning domestic confidence and adds pressure on the yuan, despite China’s recent efforts to stabilize its economy. While China has tightened capital controls, the continued capital exodus suggests that residents with resources are determined to find alternatives amid uncertain growth prospects.

China’s Agents of Chaos - Foreign Affairs

• The growing informal alliance between China, Iran, North Korea, and Russia—dubbed a new "axis of evil"—poses a significant challenge to U.S. strategic interests, even as China distances itself publicly from direct alignment with these autocracies. China, the central player due to its economic power and extensive aid capabilities, subtly supports these states while avoiding formal defense treaties, allowing it to benefit from their destabilizing actions without direct repercussions. This arrangement allows China to undermine U.S. interests while maintaining relationships with U.S. allies like Japan and the EU, who hesitate to sanction China for its tacit support of Washington's adversaries.

• Beijing’s approach could prove especially advantageous in a conflict scenario, where the combined, albeit informal, coordination among these states could stretch U.S. military resources across multiple fronts, from Asia to the Middle East and Eastern Europe. Analysts suggest that Washington must adjust its strategy by treating this alliance as an integrated threat rather than separate entities. This would mean penalizing China for the actions of its allies, thereby incentivizing Beijing to exert influence over them. Additionally, the U.S. should broaden its coalition by framing its opposition to China and its allies as a response to security threats rather than an ideological battle of democracies versus autocracies, which risks alienating potential partners in the developing world.

UK news of the week:

950-year-old Norman coins sold, marking most valuable treasure find in U.K. - The Washington Post

The coins, which date from 1066 to 1068, were found in the Chew Valley in England. They were purchased by the South West Heritage Trust for £4.3 million ($5.6 million)

Academic paper(s) of the week:

Grandmaster-Level Chess Without Search - Google DeepMind

Abstract: : This paper uses chess, a landmark planning problem in AI, to assess transformers’ performance on a planning task where memorization is futile — even at a large scale. To this end, we release ChessBench, a large-scale benchmark dataset of 10 million chess games with legal move and value annotations (15 billion data points) provided by Stockfish 16, the state-of-the-art chess engine. We train transformers with up to 270 million parameters on ChessBench via supervised learning and perform extensive ablations to assess the impact of dataset size, model size, architecture type, and different prediction targets (state-values, action-values, and behavioral cloning). Our largest models learn to predict action-values for novel boards quite accurately, implying highly non-trivial generalization. Despite performing no explicit search, our resulting chess policy solves challenging chess puzzles and achieves a surprisingly strong Lichess blitz Elo of 2895 against humans (grandmaster level). We also compare to Leela Chess Zero and AlphaZero (trained without supervision via self-play) with and without search. We show that, although a remarkably good approximation of Stockfish’s search-based algorithm can be distilled into large-scale transformers via supervised learning, perfect distillation is still beyond reach, thus making ChessBench well-suited for future research.

Influencer Dynamics - Kazimier Smith

Abstract: Consumers use social media for entertainment and to discover new products. To reach potential customers, brands pay influencers to feature products in their content. Payment depends on the size of the influencer’s audience, and the effectiveness of the endorsement relies on trust. Inauthentic recommendations may erode the relationship between an influencer and their followers. To grow their audience while maintaining its loyalty, an influencer must choose their content carefully. I develop a dynamic model in which influencers consider short and long term costs and benefits when producing content. Both organic and sponsored posts require effort to produce, and they affect the influencer’s growth. Sponsored content incurs an additional cost since it requires negotiations with brands and because the influencer may feel bad about selling out. Analyzing the career histories of 1,369 influencers, I show that organic and sponsored posts have nearly identical effects on follower growth, but influencers produce much more organic than sponsored content. The cost unique to sponsored posts generates this gap. I use observed choices to estimate the unknown cost parameters in my model, and I simulate an influencer’s optimal behavior. Influencers increase their fraction of sponsored content up to a certain size, after which it levels off. While influencers occasionally go viral, their career progression largely depends on consistent effort. Altering the return to sponsored content can affect organic content production due to dynamic incentives, but counterfactual simulations show that the effects are small. When sponsored posts become less costly in the short term, optimal production of organic posts does not change. Concerns that regulating influencer marketing reduces organic content production may be overstated.

Culture news of the week:

Painting by AI robot Ai-Da could bring more than $120,000 at Sotheby's

Sotheby’s will sell its first work credited to a humanoid robot using artificial intelligence (AI) later this month. The painting is estimated by Sotheby’s to sell for between $120,000 and $180,000 on 31 October. Fittingly, Sotheby’s will accept cryptocurrency for the transaction.

• A.I. God. Portrait of Alan Turing (2024) by Ai-Da Robot is an artist robot created by Oxford gallerist Aidan Meller.

• Meller has argued his creation is Duchampian: “Where Marcel Duchamp refused us the ability to see art in the same way as before, Ai-Da refuses us the capacity to look at the artist (and by extension the human) in the same way again,” wrote Meller and researcher Lucy Seale for last year. “What it means to be a human is changing, whether we like it or not, and this is perhaps why Ai-Da has proved so disturbing. She is reflecting this change, perhaps rather unsubtly.”

• Ai-Da paints and draws using cameras in her eyes and robotic arms. She is usually shown wearing a short, dark wig and is often in denim overalls. Critics have commented that Ai-Da is particularly beautiful, with one writing she has “mysterious hazel eyes… magnificent lips… full and puffy, like a beckoning sofa”.

Shakespeare on the wane? What does that portend?

• This is Bill Benzon;s thoughts on a recent article in the New York Times: Drew Lichtenberg, Who’s Afraid of William Shakespeare? NYTimes, Oct. 21, 2024. which notes that “In 2023-24 [in the United States], there were 40 productions of Shakespeare’s plays. There were 52 in 2022-23 and 96 in 2018-19. Over the past five years, Shakespeare’s presence on American stages has fallen a staggering 58 percent”

  • Shakespeare has long served as something of a focal point or Schelling point in the (Western) literary system, a common point of reference. Is his work being displaced from that role? Note that Lichtenberg points out, “There is a long history of theaters running from Shakespeare during times of political division or uncertainty.” Still, where does this process go? What if there is no restoration to the center? Will Shakespeare be replaced? By whom? Or is the system transforming or even dissolving?

  • What about performances in secondary schools, colleges and universities? What about performances in England, all of Britain, Western Europe, the world?

    • I will provide further thoughts on this next week when I go to see Coriolanus at the National Theatre in London

The next level of brand marketing will come from people who’ve made viral content 20 times a day, every day

• The next level of brand marketing will come from people who’ve made viral content 20 times a day, every day. Just hire a 16 y/o with 450k followers on his insta meme page, someone who catches trends early, knows the audience better, and see him help you go viral from day 1.

• Brands need to tap into this shift. it’s not just about old-school strategies anymore—it’s about staying agile, being creative, and building real content. the future of marketing is in knowing how to go viral and stay relevant.

Miscellaneous

The Nobel for Econsplaining - FT Alphaville

• This article critiques the Nobel-winning work of economists Acemoglu, Johnson, and Robinson, arguing that their institutional analysis in Why Nations Fail selectively presents the development of economic institutions in early modern England, Virginia, and Barbados as a simple dichotomy of "inclusive" versus "extractive" systems. This perspective, the author suggests, fails to capture the interconnectedness of these institutions across the Atlantic, where the profits and practices derived from slavery and colonial exploitation directly influenced British financial and industrial growth. Drawing on historiography, including the works of Edmund Morgan and Richard Sheridan, the article contends that the wealth and competencies enabling Britain's prosperity and inclusive institutions were built on and supported by extractive practices, including the brutal realities of slavery. By simplifying history to a "good vs. bad institutions" narrative, Acemoglu and Robinson's framework risks ignoring the complexities that bind inclusive institutions to exploitative systems. The article calls for economists to engage more deeply with historical analysis, suggesting that understanding the uncomfortable origins of modern institutions could inform more nuanced policy-making today.

Curious: why are ALL drivers (Uber, taxi, whatever) so right wing?

• Curious: why are ALL drivers (Uber, taxi, whatever) so right wing? I have literally never had a left wing one. And they seemingly all volunteer their political views without prompting. Is it because by driving around for a living one actually sees what’s going on in the country?

• Last three taxis I have taken this has been the case…

• Some interesting replies

  • Foreign countries = socially conservative

  • They’re sole proprietors of their own small business = fiscal conservative

  • They’ve either decided to start this as a full time job or side hustle to “pull themselves up by their bootstraps” and believe everyone else should do the same

Building the Trans-Siberian railroad cost ~15% of Russian GDP at the time! (Over ~10yrs)

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

I must give credit to a broad range of writers for pointing out many of these links, in particular Tyler Cowen of Marginal Revolution (and George Mason Univeristy)

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: ⁠⁠⁠https://substack.com/@thespecialistgeneralist⁠⁠⁠. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #5: In this week's edition of the Specialist Generalist: dive into how AI is transforming industries with Dario Amodei’s (CEO of Anthropic) bold predictions, explore exciting startups like Airship Industries reshaping air freight, and learn why AI coding assistants are revolutionising software development. Get insights into global power shifts with Saudi Arabia's tech pivot and China’s economic challenges, while Aaron Renn offers strategic career advice on climbing the right status hierarchies. Plus, discover trends in sports business, a deep dive into Palantir’s culture and how reading obituaries can inspire your path forward…. and much more!

Table of contents:

1. Top things I’ve read this week

2. AI/Technology news of the week

3. Interesting AI use cases of the week

4. A Start-Up that I like

5. Science news of the week

6. Climate news of the week

7. Sports business news of the week

8. World news of the week

9. China news of the week

10. India news of the week

11. UK news of the week

12. Academic paper(s) of the week

13. Culture of the week

14. Miscellaneous

Top things I've read this week:

Machines of Loving Grace - How AI Could Transform the World for the Better by Dario Amodei, CEO of Anthropic

• If you read one essay this week I recommend this one from Dario Amodei, largely because I think it is important that people begin to mentally prepare themselves for just how different the future (5-10 years away) might look.

• To ground the essay Dario begins by defining his understanding of powerful AI. Something that he thinks could come as early as 2026, though there are also ways it could take much longer. His focus is on the 5-10 years after that.

• Here is his definition of powerful AI:

  • ⁠⁠By powerful AI, I have in mind an AI model—likely similar to today’s LLM’s in form, though it might be based on a different architecture, might involve several interacting models, and might be trained differently—with the following properties:

    • In terms of pure intelligence⁴, it is smarter than a Nobel Prize winner across most relevant fields – biology, programming, math, engineering, writing, etc. This means it can prove unsolved mathematical theorems, write extremely good novels, write difficult codebases from scratch, etc.

    • In addition to just being a “smart thing you talk to”, it has all the “interfaces” available to a human working virtually, including text, audio, video, mouse and keyboard control, and internet access. It can engage in any actions, communications, or remote operations enabled by this interface, including taking actions on the internet, taking or giving directions to humans, ordering materials, directing experiments, watching videos, making videos, and so on. It does all of these tasks with, again, a skill exceeding that of the most capable humans in the world.

    • It does not just passively answer questions; instead, it can be given tasks that take hours, days, or weeks to complete, and then goes off and does those tasks autonomously, in the way a smart employee would, asking for clarification as necessary.

    • It does not have a physical embodiment (other than living on a computer screen), but it can control existing physical tools, robots, or laboratory equipment through a computer; in theory it could even design robots or equipment for itself to use.

    • The resources used to train the model can be repurposed to run millions of instances of it (this matches projected cluster sizes by ~2027), and the model can absorb information and generate actions at roughly 10x-100x human speed⁵. It may however be limited by the response time of the physical world or of software it interacts with.

    • Each of these million copies can act independently on unrelated tasks, or if needed can all work together in the same way humans would collaborate, perhaps with different subpopulations fine-tuned to be especially good at particular tasks.

  • We could summarize this as a “country of geniuses in a datacenter”.

• Dario believes that we need new frameworks to think about super-powerful AI in a productive way

  • Economists often talk about “factors of production”: things like labor, land, and capital. The phrase “marginal returns to labor/land/capital” captures the idea that in a given situation, a given factor may or may not be the limiting one – for example, an air force needs both planes and pilots, and hiring more pilots doesn’t help much if you’re out of planes. I believe that in the AI age, we should be talking about the marginal returns to intelligence⁷, and trying to figure out what the other factors are that are complementary to intelligence and that become limiting factors when intelligence is very high. We are not used to thinking in this way—to asking “how much does being smarter help with this task, and on what timescale?”—but it seems like the right way to conceptualize a world with very powerful AI.

• While Dario recognises that the list of positive applications of powerful AI is extremely long (it includes robotics, manufacturing, energy, and much more), he focuses on a few areas that he believes have the greatest potential to directly improve the quality of human life. I will add a few quotes for each.

  1. Biology and physical health

     1. I want to repeat this because it’s the most common misconception that comes up when I talk about AI’s ability to transform biology: I am not talking about AI as merely a tool to analyze data. In line with the definition of powerful AI at the beginning of this essay, I’m talking about using AI to perform, direct, and improve upon nearly everything biologists do.

     2. “I think their rate of discovery could be increased by 10x or more (giving us the next 50-100 years of biological progress in 5-10 years) if there were a lot more talented, creative researchers. Or, put another way, I think the returns to intelligence are high for these discoveries, and that everything else in biology and medicine mostly follows from them.”

     3. Here is a list of what Dario thinks we might expect in biology:

        • Reliable prevention and treatment of nearly all natural infectious disease.

        • Elimination of most cancer.

        • Very effective prevention and effective cures for genetic disease.

        • Prevention of Alzheimer’s.

        • Improved treatment of most other ailments.

        • Biological freedom.

        • Doubling of the human lifespan

  2. Neuroscience and mental health

     1. I think the rate of these advances will be similarly accelerated by AI and therefore that the framework of “100 years of progress in 5-10 years” applies to neuroscience in the same way it does to biology and for the same reasons

     2. It is much easier to do experiments on artificial neural networks than on real ones (the latter often requires cutting into animal brains), so interpretability may well become a tool for improving our understanding of neuroscience

  3. Economic development and poverty

     1. Dario believes that AI can help with:

        1. The Distribution of health interventions

        2. Economic growth

        3. Food security

        4. Mitigating climate change

        5. Inequality within countries

  4. Peace and governance

     1. I think of the issue as having two parts: international conflict, and the internal structure of nations. On the international side, it seems very important that democracies have the upper hand on the world stage when powerful AI is created. AI-powered authoritarianism seems too terrible to contemplate, so democracies need to be able to set the terms by which powerful AI is brought into the world, both to avoid being overpowered by authoritarians and to prevent human rights abuses within authoritarian countries.

     2. My current guess at the best way to do this is via an “entente strategy”²⁶, in which a coalition of democracies seeks to gain a clear advantage (even just a temporary one) on powerful AI by securing its supply chain, scaling quickly, and blocking or delaying adversaries’ access to key resources like chips and semiconductor equipment. This coalition would on one hand use AI to achieve robust military superiority (the stick) while at the same time offering to distribute the benefits of powerful AI (the carrot) to a wider and wider group of countries in exchange for supporting the coalition’s strategy to promote democracy (this would be a bit analogous to “Atoms for Peace”). The coalition would aim to gain the support of more and more of the world, isolating our worst adversaries and eventually putting them in a position where they are better off taking the same bargain as the rest of the world: give up competing with democracies in order to receive all the benefits and not fight a superior foe.

  5. Work and meaning

     1. As long as AI is only better at 90% of a given job, the other 10% will cause humans to become highly leveraged, increasing compensation and in fact creating a bunch of new human jobs complementing and amplifying what AI is good at, such that the “10%” expands to continue to employ almost everyone. In fact, even if AI can do 100% of things better than humans, but it remains inefficient or expensive at some tasks, or if the resource inputs to humans and AI’s are meaningfully different, then the logic of comparative advantage continues to apply. One area humans are likely to maintain a relative (or even absolute) advantage for a significant time is the physical world. Thus, I think that the human economy may continue to make sense even a little past the point where we reach “a country of geniuses in a datacenter”.

• This essay as you can imagine has generated lots of interesting commentary

  • From Matt Clancy:

    • Matt highlights the persistent divide in views on AI’s potential for rapid technological progress. Amodei suggests AI could dramatically accelerate progress, building a "country of geniuses in a data center" that achieves a century’s worth of advancement in just a few years. However, Clancy points out that even when people exchange all relevant information, their opinions on AI’s impact rarely converge. This may be due to the lack of sufficiently developed theories to rigorously predict outcomes, leading people to rely heavily on their personal intuitions. For example contrast economists, whose experiences with productivity gaps and views on slow tech diffusion are shaped by historical examples such as the delayed impact of computers and the frictions seen in global economic development, vs. technologists like Amodei who are more comfortable forecasting rapid technological leaps. Ultimately, in such a novel and unpredictable scenario as the rise of AI, it’s unclear whose intuitions—economists' or technologists'—are more reliable

  • From Greer on X:

    • Dario spends a large portion of his essay talking about the need for a liberal-democratic political coalition to control the commanding heights of AI technology to keep this technology from authoritarian powers. At the same time he proposes AGI aided governance of much of the third world, guiding industrial policy there to supercharge global GDP growth. It will be hard to keep the commanding heights in (western) liberal hands while simultaneously handing the technology over to the governments of the developing world. These goals are probably not compatible—it will take a lot of work to make them so.

  • From Tyler Cowen:

    • I view human imperfections, and current institutional and legal constraints as more binding than Dario does, and thus I think speeds of progress will be lower than he does. But there is much in his essay I agree with.

Reflections on Palantir (A retrospective of an eight-year stint) by Nabeel S. Qureshi

• This article reflects on an eight-year tenure at Palantir, highlighting the company's unique culture, business model, and evolution.

• Per Tyler Cowen: as Nabeel and a few others have noted, there should be many more pieces trying to communicate what various businesses and institutions really are like.

• Here are some key excerpts that I found notable from the various sub-chapters:

  • Why I joined:

    • The combo of intellectual grandiosity and intense competitiveness was a perfect fit for me. It’s still hard to find today, in fact – many people have copied the ‘hardcore’ working culture and the ‘this is the Marines’ vibe, but few have the intellectual atmosphere, the sense of being involved in a rich set of ideas. This is hard to LARP – your founders and early employees have to be genuinely interesting intellectual thinkers. The main companies that come to mind which have nailed this combination today are OpenAI and Anthropic. It’s no surprise they’re talent magnets

  • Notes on Culture:

    • ⁠Good founders have an instinct for reading rooms, group dynamics, and power. This isn’t usually talked about, but it’s critical: founding a successful company is about taking part in negotiation after negotiation after negotiation, and winning (on net). Hiring, sales, fundraising are all negotiations at their core. It’s hard to be great at negotiating without having these instincts for human behavior. This is something Palantir teaches FDEs, and is hard to learn at other Valley companies.

    • ⁠⁠One of my favorite insights from Tyler Cowen’s book ‘Talent’ is that the most talented people tend to develop their own vocabularies and memes, and these serve as entry points to a whole intellectual world constructed by that person. Tyler himself is of course a great example of this. Any MR reader can name 10+ Tylerisms instantly - ‘model this’, ‘context is that which is scarce’, ‘solve for the equilibrium’, ‘the great stagnation’ are all examples. You can find others who are great at this. Thiel is one. Elon is another (“multiplanetary species”, “preserving the light of consciousness”, etc. are all memes). Trump, Yudkowsky, gwern, SSC, Paul Graham, all of them regularly coin memes. It turns out that this is a good proxy for impact. This insight goes for companies, too, and Palantir had its own, vast set of terms, some of which are obscure enough that “what does Palantir actually do?” became a meme online. ‘Ontology’ is an old one, but then there is ‘impl’, ‘artist’s colony’, ‘compounding’, ‘the 36 chambers’, ‘dots’, ‘metabolizing pain’, ‘gamma radiation’, and so on. The point isn’t to explain all of these terms, each of which compresses a whole set of rich insights; it’s that when you’re looking for companies to join, you could do worse than look for a rich internal language or vocabulary that helps you think about things in a more interesting way.

  • Bat signals

    • ⁠⁠Another thing I can trace back to Peter is the idea of talent bat-signals. Having started my own company now (in stealth for the moment), I appreciate this a lot more: recruiting good people is hard, and you need a differentiated source of talent. If you’re just competing against Facebook/Google for the same set of Stanford CS grads every year, you’re going to lose. That means you need a set of talent that is (a) interested in joining you in particular, over other companies (b) a way of reaching them at scale. Palantir had several differentiated sources of recruiting alpha.⁠⁠

  • Morality:

    • Break down the company’s work into three buckets – these categories aren’t perfect, but bear with me:

      1. Morally neutral. Normal corporate work, e.g. FedEx, CVS, finance companies, tech companies, and so on. Some people might have a problem with it, but on the whole people feel fine about these things.

      2. Unambiguously good. For example, anti-pandemic response with the CDC; anti-child pornography work with NCMEC; and so on. Most people would agree these are good things to work on.

      3. Grey areas. By this I mean I mean ‘involve morally thorny, difficult decisions’: examples include health insurance, immigration enforcement, oil companies, the military, spy agencies, police/crime, and so on.

    • Every engineer faces a choice: you can work on things like Google search or the Facebook news feed, all of which seem like marginally good things and basically fall into category 1. You can also go work on category 2 things like GiveDirectly or OpenPhilanthropy or whatever.

    • The critical case against Palantir seemed to be something like “you shouldn’t work on category 3 things, because sometimes this involves making morally bad decisions”. An example was immigration enforcement during 2016-2020, aspects of which many people were uncomfortable with. But it seems to me that ignoring category 3 entirely, and just disengaging with it, is also an abdication of responsibility. Institutions in category 3 need to exist. The USA is defended by people with guns. The police have to enforce crime, and - in my experience - even people who are morally uncomfortable with some aspects of policing are quick to call the police if their own home has been robbed. Oil companies have to provide energy. Health insurers have to make difficult decisions all the time. Yes, there are unsavory aspects to all of these things. But do we just disengage from all of these institutions entirely, and let them sort themselves out?

  • What’s next:

    • Do I remain bullish on the company? Yes. ⁠⁠The big productivity gains of this AI cycle are going to come when AI starts providing leverage to the large companies and businesses of this era - in industries like manufacturing, defense, logistics, healthcare and more. Palantir has spent a decade working with these companies. AI agents will eventually drive many core business workflows, and these agents will rely on read/write access to critical business data. Spending a decade integrating enterprise data is the critical foundation for deploying AI to the enterprise. The opportunity is massive⁠⁠.

AI/Technology news of the week:

Generative AI’s Act o1 - Sequoia Capital

Two years into the Generative AI revolution, research is progressing the field from “thinking fast”—rapid-fire pre-trained responses—to “thinking slow”— reasoning at inference time. This evolution is unlocking a new cohort of agentic applications.

• Here is a summary:

  • The foundational layer of the Generative AI market is stabilizing with major players like Microsoft/OpenAI, AWS/Anthropic, Meta, and Google/DeepMind. With next-token predictions becoming cheaper and more accessible, the focus is now on enhancing AI’s reasoning capabilities, such as OpenAI's “Strawberry” (o1) model, which uses inference-time compute for reasoning, akin to AlphaGo's deliberate game strategies.

  • This evolution introduces "System 2" thinking—AI models that not only mimic patterns but also generate, evaluate, and choose the best outcomes in real-time. It’s a significant leap from System 1's instinctual responses to deeper problem-solving, essential for complex tasks like coding, math, and science.

  • As inference-time compute becomes central to AI's reasoning abilities, there’s a shift towards "inference clouds," which dynamically scale based on task complexity. This brings forward a new wave of AI applications with custom cognitive architectures, designed to perform specific tasks rather than generalized outputs.

  • The emerging applications, such as AI legal assistants, medical scribes, and software engineers, are transforming entire sectors. These agentic applications sell outcomes rather than software, opening new markets by reducing labor costs and expanding service delivery.

  • For startups, the application layer holds the most promise, where cognitive architectures bridge AI capabilities with real-world workflows. As AI advances, reasoning models will become more refined, driving the next wave of innovation, with potential for multi-agent systems working in tandem.

  • Ultimately, the industry is awaiting its "Move 37" moment—a breakthrough where AI surprises us with superhuman reasoning, marking a significant milestone towards AGI (Artificial General Intelligence).

• A couple of excerpts I found particularly interesting:

  • Service-as-a-Software: The cloud transition was software-as-a-service. Software companies became cloud service providers. This was a $350B opportunity. Thanks to agentic reasoning, the AI transition is service-as-a-software. Software companies turn labor into software. That means the addressable market is not the software market, but the services market measured in the trillions of dollars.

    • What does it mean to sell work? Sierra is a good example. B2C companies put Sierra on their website to talk with customers. The job-to-be-done is to resolve a customer issue. Sierra gets paid per resolution. There is no such thing as “a seat”. You have a job to be done. Sierra does it. They get paid accordingly.

  • A New Cohort of Agentic Applications:

    • These companies look different than their cloud predecessors:

      • Cloud companies targeted the software profit pool. AI companies target the services profit pool.

      • Cloud companies sold software ($ / seat). AI companies sell work ($ / outcome)

      • Cloud companies liked to go bottoms-up, with frictionless distribution. AI companies are increasingly going top-down, with high-touch, high-trust delivery models.

    • Here are some examples.

      • Harvey: AI lawyer

      • Glean: AI work assistant

      • Factory: AI software engineer

      • Abridge: AI medical scribe

      • XBOW: AI pentester

      • Sierra: AI customer support agent

  • The investment universe:

    • The most interesting layer for venture capital. ~20 application layer companies with $1Bn+ in revenue were created during the cloud transition, another ~20 were created during the mobile transition, and we suspect the same will be true here.

Software Explosion - Philip Joubert

Software has eaten the world. Now AI will digest it.

• Some of the areas where I’m expecting a software explosion:

  1. Replacing SaaS with in-house software: Why pay SaaS-vendors like Salesforce millions when you can build your own CRM, ATS or HR system that’s customized to your business?

  2. Proliferation of micro software: There are thousands of small points of friction within companies that could be streamlined or automated with some code and sticky tape.

  3. Expansion of product offerings: AI raises the bar for customer expectations. We’re all going to need to build far broader, smarter products.

  4. Automation of human tasks: The number of jobs that can be automated completely or partially will increase dramatically - all of which requires us to build a lot more software.

  5. Tech-enabled becomes the default: Building software used to be reserved for large companies and funded startups, soon every company will be able to afford to create their own software.

2024: The Year AI Agents Matured Beyond Single Foundation Models

• 2024 marks a pivotal year for AI agents, with new frameworks like AutoGen, Crew AI, LangGraph, and OpenAI’s Swarm enabling multiple AI agents to collaborate on complex tasks. These agents go beyond static LLMs by acting autonomously, planning, and executing across domains, which is driving innovation across industries. AI agents can now handle scalability, complexity, and cross-system integration, democratizing AI development and transforming sectors like finance and healthcare. Challenges remain, such as computational costs, coordination issues, and the need for human oversight, but agent-based AI is gaining traction, with $8.2 billion in funding in 2024 alone. AI agents are expected to reshape workflows, making businesses more efficient by automating complex, multi-step tasks.

• In late September, early-stage VC firm Felicis released a market map showing some of the early leaders in the AI agent economy with the public market incumbents they’re trying to disrupt:

Offshoring and AI Agents - The Diff

• An excerpt:

  • A 99% reliable system can actually be safer than a 99.999% reliable one, because when errors are sufficiently rare, it's hard to pay close enough attention to spot them. This is especially true if the failure isn’t obvious to another human in the first place, and the way AI assistants work means most of their errors are either glaringly obvious to a human being, or subtle enough to get past a human. Finding these errors, by definition, is review-intensive, especially in a distributed system, where a perfectly logical line of code can, in the right context, create a problem.

• Here is a summary of the article:

  • AI coding assistants increase productivity, but they also require more code review and understanding of complex systems. As developers rely on AI to handle unfamiliar tasks, the quality of work depends on clear problem-solving and oversight. While AI makes code generation faster, it introduces challenges in reviewing and understanding machine-written code, especially in complex systems.

  • Ultimately, AI is reshaping software development, with more emphasis on specification and verification, as developers increasingly rely on assistants to automate tasks. The future may involve more reviewing and less coding, with AI playing a growing role in the industry’s shift towards automation.

What has surprised you most this year? From 5 questions for Rohit Krishnan - Politico

• The development of decentralized training for AI models, from DiLoCo and DiPaCo from Google DeepMind to Distro coming up from Nous Research. It completely changes the game in terms of how we ought to think about large models and what we can do to train them. This means pure compute thresholds are not going to be very useful, and that we will have even less of a way to centrally control the means of knowledge production.

  • Traditionally, training large AI models requires enormous amounts of computational power from centralized servers, typically owned by big companies like Google, Microsoft, or OpenAI. Decentralized training changes this by allowing different parts of an AI model to be trained in separate locations by different computers or groups, and then combining the results. Instead of needing one massive supercomputer or central system, decentralized training spreads the work across many smaller systems, which could be geographically scattered and independently owned.

    • Why is this important?

      1. Lower Barriers to Entry: This approach breaks the dependency on supercomputers or large centralized systems, meaning more people or groups (like small companies, academic institutions, or even individuals) can contribute to building powerful AI systems.

      2. Democratization of AI: In a centralized system, a few companies control the development and deployment of the most powerful AI models. Decentralized training distributes this power more widely, making it harder for a single company or government to dominate AI development. This could lead to more innovation in AI research.

      3. More Efficient Use of Resources: Decentralized systems can potentially make better use of available resources by leveraging unused computing power across different devices or locations. This could improve the speed and efficiency of AI model training.

Sequoia Capital and the evolution of the VC industry | FT Film

• Everybody, including Sequoia, is rushing to find the next big winner out of AI. 60% of Sequoia's current investments are AI.

• SEBASTIAN MALLABY: Artificial intelligence is at a point where, let's say, the mobile smartphone ecosystem was around 2007-2008. So the iPhone came out in 2007. The App Store was created in 2008. WhatsApp and Instagram didn't come until 2009 and 2010. So it always takes a little while after the new platform is created before you get the new startups that are going to go to more than a $1 dollar valuation based on this new opportunity. And that's where we are today with AI.

This $10M U.S. Army Laser Melts Drones With $3 Beams | WSJ Equipped

• This is important because right now in an age of modern drone warfare we need cost-effective alternatives to missiles. The laser of BlueHalo profiled here can choot down a drone at $3 cost per shot compared to the multi-million-dollar missile strikes traditionally used. It uses AI and fiber optics to track and melt drones within a three-mile range. While still in its early stages, laser technology is evolving, and companies like BlueHalo are working on improving their effectiveness, weight, and ease of maintenance.

Boston Dynamics and Toyota Research Institute Announce Partnership to Advance Robotics Research

• Boston Dynamics and Toyota Research Institute (TRI) announced today they will join forces, combining two of the world’s leaders in artificial intelligence and robotics. The research partnership aims to accelerate the development of general-purpose humanoid robots utilizing TRI’s Large Behavior Models and Boston Dynamics’ Atlas robot.

Watch SpaceX Starship's epic 5th launch, catch, re-entry and splashdown in time-lapse

• Goes without saying this is an incredible achievement and brings us ever close towards inter-planetary travel

  • More on this in Elon Dreams and Bitter Lessons - Stratechery

TikTok execs know exactly how many videos it takes to get addicted to its platform

• According to TikTok's own internal research, users only need to watch 260 videos before they could become addicted to the app. That same research found that "compulsive usage correlates with a slew of negative mental health effects like loss of analytical skills, memory formation, contextual thinking, conversational depth, empathy, and increased anxiety." It acknowledged that the algorithm had "better engagement" with young people.

• I think there is a strong marginal benefit to being on TikTok (largely from a cultural perspective - understanding trends) but am I also addicted - maybe?

Can AI help Africa close the development gap?

• AI holds significant potential to drive transformative development in Africa. Here are some areas the article touches upon:

  1. Healthcare Innovation: AI is making healthcare more accessible and affordable in regions where resources are limited

  2. Empowering Farmers: AI is helping African farmers improve crop yields and combat diseases.

  3. Tailored Education: AI is transforming education by providing personalized learning experiences. In Ethiopia, AI-driven tutoring programs like SkillBridge are being developed to teach students in their native languages, such as Amharic and Afan Oromo.

  4. Improving Accessibility: AI-powered tools, such as Google’s Project Relate, are giving individuals with disabilities new opportunities to communicate and integrate into the workforce.

  5. Boosting Climate Resilience: Google’s AI research in Accra is helping to monitor flash floods and predict weather events in areas where traditional weather monitoring systems are scarce.

  6. Youth-Driven Development: With Africa's median age being 19, AI presents a unique opportunity to empower the continent’s youth. As more young Africans develop AI skills through initiatives like A2SV (Africa to Silicon Valley), they can harness AI to drive homegrown solutions to local challenges..

Where the world's largest hedge fund sees investing opportunities - Briefings from Goldman Sachs

• Karen Karniol-Tambour, the co-chief investment officer of Bridgewater Associates:  “All it will take is one AI-powered startup in your industry to say: 'Wait a minute, I better go invest a lot in AI as well… You can look back to the 1990s and that midcycle easing, and how big the capex cycle in the internet boom was in powering the next leg. I think this is about as good of a backdrop as you could get so long into an expansion to be able to sustain growth. And you have a Fed that, with inflation where it is, is both willing and able to do more if it needs to.” 

Sam Altman’s Worldcoin Gets a New Name, and New Hardware

“Humanity now faces a new existential question: How will humans & AI coexist?” the company said in a statement. 

• Worldcoin, a startup co-founded by OpenAI Chief Executive Officer Sam Altman, announced a new name and introduced the latest version of its eyeball-scanning Orb at an event in San Francisco on Thursday. The company will be called World, and will use the new Orb to provide identification services — a task that the company said will be increasingly critical as artificial intelligence-enabled deepfakes become more adept at mimicking people. 

• Orb scans people’s eyes in order to provide biometric identification. The company’s latest model will use Nvidia Corp.’s Jetson chipset and will be rolling out the ability for people to get the machine delivered to their house for real-time identification needs. The delivery will be like ordering a pizza, Chief Device Officer Rich Heley said on stage Thursday.  The ID service, called Deep Face, will address the problem of AI deepfakes and offer a “powerful new way to combat fraud,” the company said in the statement.

Interesting AI use cases of the week:

Microsoft's Mustafa Suleyman says his team is crafting AI companions who will see and remember everything we do and which will constitute an intimate relationship with AI

AI Will Transform Philanthropy, Too - Tyler Cowen

• AI will enable smaller groups to direct larger impact:

  • One big change is that AI will enable individuals, or very small groups, to run large projects. By directing AIs, they will be able to create entire think tanks, research centers or businesses

• How AI should impact philanthropic hiring:

  • The productivity of small groups of people who are very good at directing AIs will go up by an order of magnitude. Philanthropists ought to consider giving more support to such people. Of course that is difficult, because right now there are no simple or obvious ways to measure those skills. But that is precisely why philanthropy might play a useful role. More commercially oriented businesses may shy away from making such investments, both because of risk and because the returns are uncertain. Philanthropists do not have such financial requirements.

• How AI might reframe philanthropic priorities

  • Strong AI capabilities also mean that the world might be much better over some very long time horizon, say 40 years hence. Perhaps there will be amazing new medicines that otherwise would not have come to pass, and as a result people might live 10 years longer. That increases the return — today — to fixing childhood maladies that are hard to reverse. One example would be lead poisoning in children, which can lead to permanent intellectual deficits. Another would be malnutrition. Addressing those problems was already a very good investment, but the brighter the world’s future looks, and the better the prospects for our health, the higher those returns.

  • The flip side is that reversible problems should probably decline in importance. If we can fix a particular problem today for $10 billion, maybe in 10 years’ time — due to AI — we will be able to fix it for a mere $5 billion. So it will become more important to figure out which problems are truly irreversible. Philanthropists ought to be focused on long time horizons anyway, so they need not be too concerned about how long it will take AI to make our world a fundamentally different place.

From Elon Musk to cop car chases, how a software engineer launched a police AI startup - Tech Crunch

• The company, Abel, creates AI that uses body cam footage and dispatch call data to fill out normally time-consuming police reports. On Thursday, it announced Abel has raised a $5 million seed round led by Day One Ventures with participation from Long Journey Ventures and backing from Y Combinator as part of the summer 2024 class.

OpenAI Startup Fund Led $9 Million Round In Accounting AI Agent Startup

• Kick, a Palo Alto, CA-based startup that creates accounting and bookkeeping AI agents, raised $9M in funding. The round was led by General Catalyst and OpenAI Startup Fund, with support from angels likes Lachy Groom (Ex-Stripe), Michael Vaughn (Ex-Venmo), Eric Glyman, Karim Atiyeh (Founders, Ramp), Nick Davidov (DVC), Sahil Lavingia (CEO, Gumroad), Austen Allred (CEO, Bloomtech), Tikhon Berstram (Co-Founder, Parse), Soleio Cuervo (Ex-Facebook), Scott Belsky (CSO, Adobe) Joshua Browder (CEO, DoNotPay) and 50 others.

Interesting Start-Up news of the week:

Harry Stebbings, U.K. podcaster turned VC, has closed a $400 million fund for 20VC, focusing on European startups. $125 million will target seed investments, and $275 million for Series A rounds. Stebbings raised the fund in four weeks, backed by U.S. investors and institutional money.

PE plumbers (Matt Levine)

• In his Money Stuff blog, Matt posits that "private equity" is often essentially a sophisticated term for managing everyday businesses like plumbing or HVAC services, industries that private equity firms are increasingly acquiring. Historically, family-owned businesses would pass down through generations, with the next in line learning modern management techniques to maintain or grow the company. However, today’s model sees private equity firms stepping in, purchasing these businesses, and applying their financial expertise to create more efficient operations, offering large paydays to previous owners. This shift has transformed industries like plumbing and HVAC into lucrative opportunities for private equity, with nearly 800 companies acquired since 2022.

  • Given the surging investor interest, [Alpine Investors founder Graham] Weaver says, anyone with entrepreneurial ambitions should take a second look at the trades, which offer steady income via unclogging toilets, fixing boilers and installing new air-conditioning units. “You can build a business that’s going to be worth $10-30 million and have a ready list of buyers to sell it to,” he says. “Ten years ago, there was no one to sell it to.”

Some of my notes from the Art of Hiring event earlier today - Cory Levy

• @rabois

  • - don't hire anyone over the age of 30 - recruiters are probably ineffective but still important b/c they drive hiring momentum - it takes 3-30 mins to evaluate someone (probably 3 though) - undiscovered talent is mispriced. interns can be MVPs on teams - people who positively manipulate the world around them are great hires

• @bchesky

  • - goal is to have smallest team possible (i.e. be the Navy SEALs... not the Navy) - be careful of B players. The old saying "A players hire A players B players hire C players" should go like "A players hire A players....and B players hire LOTS of C players b/c they are so bad" - be in the details. so much so that people complain. and then be in even more details. - when hiring someone, they are guilty until proven innocent. someone should be selling their case on why to join.

• @eglyman

  • - looks for "aliens" and/or exceptional candidates with little competition - ranks candidates "strong no, no, yes, strong yes". won't hire people who have all yeses. - recommended to read great books on this like: Wooden on Leadership, Insanely Simple: The Obsession That Drives Apple's Success, High Output Management, and The Score Takes Care of Itself

• @mike_sheb

  • - be a broken record. constantly repeat why you are here (mission, goals, etc.) - sweat the small stuff - ask candidates if they are in the phase of their lives where they can work like an olympian to help screen people out.

A Start-Up that I like:

Airship Industries

• Airship Industries is designing large, rigid-body cargo airships aimed at transforming the transoceanic air freight market.

• The company's key innovation lies in its ability to significantly reduce end-to-end delivery times compared to current air freight services, while lowering handling costs, delays, and breakage.

• Its airships promise to operate at 75% lower fuel consumption and carbon emissions, without needing breakthroughs in sustainable fuel. Unlike traditional air freight, Airship Industries' vehicles can offer faster, more efficient delivery and capture higher margins due to reduced operational costs and lower fuel dependency. Their unique selling point is their ability to directly compete with 747 freighters on delivery speed, while being far more profitable, sustainable, and resilient against fuel price volatility.

On the topic of aviation I’ll also add another Start-Up here:

Astro Mechanica

• Recently backed in the YC W24 batcyh, the team at Astro Mechanica are attempting to build a new jet engine that can enable supersonic flight in a sustainable fuel efficient way

• A new kind of engine, to power a new kind of plane, to enable a wild new future of air travel.

Science news of the week:

Patrick Hsu: A Trailblazer in Digital Biology - At the Interface of biomedical science and A.I.

• This is a great podcast between Eric Topol and Patrcik Hsu, co-founder and core investigator at the Arc Institute, and who recently led the team that discovered bridge RNAs, which may be considered CRISPR 3.0 for genome editing. He is currently building new generative A.I. models for life science.

• I recommend listening / reading the full transcript but here are some key takeaways:

  • Arc Institute’s Vision and Creation: Patrick Hsu, along with Patrick Collison and Silvana Konermann, created Arc Institute to bridge biomedical science and machine learning. Its innovative model eliminates the typical burdens of grant writing for investigators, encourages large-scale collaborative projects, and aims to translate scientific discoveries into real-world applications. Hsu highlights that new organizational models, combining elements from academia and industry, are crucial for achieving transformative scientific breakthroughs. Arc's model itself includes core investigators who operate across partner universities but are physically located at Arc's headquarters. Additionally, innovation investigators receive unrestricted funding for exploratory research & technical staff are hired to focus on R&D initiatives.

  • Groundbreaking Discovery of Bridge RNAs: The discovery of Bridge RNAs represents a significant leap in genome editing beyond traditional CRISPR. This new mechanism allows for precision editing without DNA cuts, marking a potential shift towards more efficient, scalable genomic alterations. The discovery showcases Arc's commitment to pioneering technologies that push the boundaries of current biological capabilities.

  • Digital Biology and the Future of Life Sciences: Hsu aligns with Jensen Huang’s prediction that biology is transforming from a science to an engineering discipline. Arc is at the forefront of this shift, using AI models to better understand and design biological systems. Hsu anticipates that AI-driven models that integrate understanding across molecular, cellular, and organismal scales will transform biology into a predictive discipline, offering unprecedented insight into biological processes and accelerating drug discovery.

  • Challenges with Computing Power and Open-Source Models: Hsu argues that academic sciences are unprepared for the level of resources that are required for doing this type of cutting edge computational work. There are top Stanford computer science professors or computational researchers who have a single GPU in their office, and that's actually what their whole lab runs off of. Arc has overcome this by building private GPU clouds.

  • AI Scientist Concept and Automation in Life Sciences: While AI holds great potential as an autonomous research assistant, Hsu believes human scientists will remain essential for creativity and innovation. AI can augment scientific research, but the human element—particularly in hypothesis generation and experiment design—remains critical for major scientific breakthroughs.

Obesity is finally falling (from At least five interesting things to start your week (#51)) - Noahpinion

• Recent data from the U.S. National Health and Nutrition Examination Survey (NHANES) suggests that, after decades of rising obesity rates, America may have reached "peak obesity," with rates beginning to fall since 2020. This trend coincides with the rise in the use of weight-loss drugs like Ozempic, which has become increasingly prescribed since 2018, especially for diabetes, a condition closely linked with obesity. Interestingly, the decline is most pronounced among college-educated Americans, who are more likely to access new medical treatments, suggesting a multi-causal factor involving both technological advances and potentially cultural shifts. While this marks a significant success for medical intervention, it also highlights a broader idea: that human health problems like obesity, often treated as personal failings, can be reframed as technological or medical challenges. This shift in perspective invites consideration of how other health issues, such as addiction, might similarly benefit from technological solutions rather than societal stigma.

US startup charging couples to ‘screen embryos for IQ’

• Here is a summary: A US startup, Heliospect Genomics, is offering wealthy couples the controversial option to screen embryos for IQ and other traits like height and risk of obesity, using advanced genetic technology. Charging up to $50,000 for analyzing 100 embryos, Heliospect claims its methods can increase a child’s IQ by six points. While legal in the US, this technology raises significant ethical concerns, including fears of reinforcing inequality and the normalization of "superior" genetics. Critics argue that such genetic selection could widen social divides and promote dangerous eugenic ideas. Heliospect uses genetic data from the UK Biobank, which has sparked debate about the ethical use of public genetic resources for commercial purposes. The startup's services are in "stealth mode" before an official launch, and the company insists it operates within legal boundaries. Bioethicists warn this development represents an ethical "minefield," calling for deeper public engagement on the issue.

• If this is a topic that interests you, there are few people more knowledge on the topic than Steve Hsu, Theoretical Physics and of Computational Mathematics, Science, and Engineering at Michigan State University and founder of the company Genomic Prediction

  • I remember him discussing this topic back in 2018 where mentions a scenario where an IVF physician is advising parents who have 2 viable embroyos ready for implantation: Do you inform them about the polygenic score difference between the embryos? Note, in the very near future this question will no longer be hypothetical..

  • This question is becoming less hypothetical and more real! And we can’t shy away from this topic for much longer…

    • Here is the link to his discussion with Dwarkesh: Steve Hsu - Intelligence, Embryo Selection, & The Future of Humanity

Climate/Energy news of the week:

Google: New nuclear clean energy agreement with Kairos Power

• Since pioneering the first corporate purchase agreements for renewable electricity over a decade ago, Google has played a pivotal role in accelerating clean energy solutions, including the next generation of advanced clean technologies. Today, we’re building on these efforts by signing the world’s first corporate agreement to purchase nuclear energy from multiple small modular reactors (SMRs) to be developed by Kairos Power. The initial phase of work is intended to bring Kairos Power’s first SMR online quickly and safely by 2030, followed by additional reactor deployments through 2035. Overall, this deal will enable up to 500 MW of new 24/7 carbon-free power to U.S. electricity grids and help more communities benefit from clean and affordable nuclear power.

• Although as Zvi notes: That is great and is fast by nuclear power standards, and also slower than many people’s timelines for AGI.

EVs are just going to win - Noahpinion

• Here is a summary: The narrative that the electric vehicle (EV) revolution was stalling has collapsed, as EV sales rebounded in the U.S., setting new market share records. Globally, EV sales are growing rapidly, and while only a small percentage of the global vehicle fleet is electrified, fundamental advantages suggest EVs will soon dominate. Key drivers include falling battery costs, set to reach price parity with internal combustion engine (ICE) vehicles by 2026, and increased convenience, with most EV owners able to charge at home. EVs require less maintenance, offer faster acceleration, and are quieter than combustion cars. As EV adoption rises, gas stations may become less profitable, further driving the shift. While the U.S. has lagged behind China in EV production, the global transition to EVs is inevitable and will benefit both consumers and the environment.

Sports business news of the week:

Changing the Game - Goldman Sachs

The global sports ecosystem has been underpriced and misunderstood for years.

Over the past decade, the sports landscape has undergone a holistic transformation as heightened consumer interest, technological advances, and a diversified media engine have created a dynamic sports ecosystem that is integrated into our culture like never before. Investors deploying capital position themselves at the nexus of strategic decision-making, with the opportunity to shape the future of sports.

• I thought this chart was particularly interesting. As media rights are expected to flatline, sports rights holders are focusing (and are going to have to focus) on other verticals to continue growing revenue.

• Aaron Miller of Will Ventures makes the following predictions:

  • There's more media than ever competing for our attention, from YouTube to Tik Tok to Roblox. As that trend accelerates, attention will further consolidate around the most culturally-relevant leagues like the NFL, NBA & EPL.

  • The other major leagues will innovate. This season the MLB addressed its decreasing audience by making key changes (e.g. the pitch clock), and postseason viewership is now crushing it. NHL and smaller European soccer leagues will similarly innovate on their event and content formats to keep their media packages valuable.

  • A fully new crop of sports leagues will emerge. TMRW Sports, The Snow League, Kings League and other emerging leagues will lean into digital-first strategies, capture large fanbases, and be on this chart in ten years.

Competitive socialising pulls Britons from the pub - The FT

• The UK has seen a boom in venues focused on competitive socializing, combining games like cricket, darts, and golf with food and drink, as corporate and social events move away from alcohol-centric gatherings. Venues like Sixes Social Cricket, which combines real cricket balls with virtual bowling, are part of this trend. These venues have grown internationally, with brands like Puttshack, Flight Club, and Lane7 expanding in Europe and beyond. The number of UK competitive socializing venues has surged by 40% since 2018, and is expected to keep rising, fueled by shifting consumer preferences toward experiences over goods, health-consciousness, and the need for team-building activities post-Covid. However, with increasing competition, some consolidation and fading of certain trends, like virtual reality, are anticipated.

  • Also this week: Cheque mates: Harry Kane and Magnus Carlsen top up stakes in football brand

    • England captain Harry Kane and chess grandmaster Magnus Carlsen have taken part in a $100m funding round for the company behind the football-themed entertainment and dining venue Toca Social.

    • The move will help Toca Football launch further sites in London, where it already has a venue at the O2, as well as Paris, Mexico and the US, and invest in new technology for its Toca Soccer training centres, the company said.

World news of the week:

Just a reminder that Africa is the world’s future

• Cities have been magical engines of growth throughout human history. But the link btw urbanization & progress is breaking down in African cities today. At the same time, Africa is adding just under 1 billion new urbanites to its cities btw now and 2050.

• There are many bottlenecks to seizing this urban future. One crucial area is the lack of capacity of cities and towns across the continent to adequately grapple with exploding urban growth. The UK, for example, has ~37 urban planners per 100k ppl. The US has ~12 per 100k. In Africa, it's about 0.89 per 100,000 ppl. In 2040, African cities will need about 600,000 engineering graduates per year to design, build, and service its cities.

• The @AfricaUrbanLab aims to train the city builders and urban leaders of tomorrow to harness rapid urbanisation for human prosperity.

Back to the future: Russia to unveil new Stalin statue

• The current culture that is Russia.

  • There was a public demand to commemorate the dictator responsible for millions of deaths, local politician says.

  • The governor says that “with all understanding of the ambiguous interpretation of the role” of Stalin, Russians must recognize his “great achievements” and not only know their country’s history, but “honor and be proud of it.”

  • Russian history is “a single and indivisible chain of interconnected, interlocking links of the historical process, each of which has shaped the strength, spirit and will of our great nation,” he said

    • Whilst the book was published before the invasion of Ukraine, I’d highly reccommend reading Putin's People: How the KGB Took Back Russia and Then Took on the West by Catherine Belton

The Problem Isn't Public R&D by Pieter Garicano

• Unlike what is commonly believed, Europe is not lagging in government investment in science. As a share of GDP, Europe spends roughly 0.74% on public sector R&D, compared to the United States' 0.69%. Total governmental R&D spending in 2021 is roughly 131 billion in the US and 108 billion in the EU. The actual R&D gap is in private sector spending, where Europe spends 1.3% of GDP compared to the United States' 2.4%. That gap is good for 341 billion in R&D spending in 2021.

• Indeed,  a common misconception among some economists and policy makers is that the United States leads in innovation due to the work done by publicly funded researchers, particularly by the Defense Department. In reality, Defense Department R&D represents less than 10% of total R&D in the U.S. Many of the great innovations to come out of the Pentagon are now decades old. The areas where Europe is now most clearly falling behind are almost wholly being driven by the private sector:  AI, space, EVs.

  

Saudi construction project of the day

• Saudi Arabia is preparing to begin construction work on its next giga-project: a cube-shaped skyscraper big enough to fit 20 Empire State Buildings.

• The Mukaab will be 400-meters on each side when construction is finished, which would make it the largest built structure in the world. The building will be the centerpiece of New Murabba, a community the country hopes will be a new destination within the capital city of Riyadh.

• “Ultimately, a capital city the size of Riyadh deserves to have a global, central icon as other capital cities do.”

  • Also Saudi Arabia this week:

    • ⁠⁠The Neom giga-project in Saudi Arabia is currently using one fifth of all the steel produced in the world, an official said on Monday. The futuristic city will be the world’s largest customer for construction materials for several decades, said Manar Al Moneef, Neom’s chief investment officer. She told the Global Logistics Forum in the King Abdullah Financial District in Riyadh that the $500 billion project would be one of the world’s leading drivers of the global logistics sector in coming years. “Neom is going to be the largest customer over the next decade. If you look at our demand in logistics it’s 5 percent of the global logistics market,” she told the forum, in rare public comments.

China news of the week:

Why did China’s police chief make a rare foray into economic policy?

• China's top police officer, Wang Xiaohong, made a rare comment on economic matters, emphasizing the importance of risk management and stability during a visit to Tianjin. His remarks, delivered in his capacity as a senior Communist Party member and state councillor rather than public security minister, suggest growing concern among officials about meeting this year's 5% economic growth target. With China achieving only 4.7% GDP growth in the second quarter, Wang's comments echo President Xi Jinping’s recent calls for intensified efforts to meet the economic targets amid signs of a faltering economy.

China’s economic goals (from At least five interesting things to start your week (#51)) - Noahpinion

• Noah’s mental model of Xi Jinping since at least 2021:

  1. Xi’s overriding priority is to make China a geopolitically dominant and militarily strong nation.

  2. Xi sees certain types of economic activity — especially advanced manufacturing and agriculture — as crucial for national power.

  3. Xi sees other types of economic activity — real estate development, health care, internet software, finance, private education, etc. — as less useful for national power.

  4. Thus, Xi will prioritize the growth of advanced manufacturing above all else. His attitude toward the other sectors will depend on whether he thinks, on balance, that they help build political support for his national greatness project, or whether he thinks they divert resources from that project.

• Noah however, argues that China’s recent pivot toward stimulus (discussed in a previous edition of the Specialist Generalist) represents a shift in Xi’s thinking. In 2021-23 he thought that diverting resources from real estate, IT, and services was crucial for boosting advanced manufacturing. But now, with a demand shortage hurting the country’s economy, and many Chinese people becoming disaffected, he has probably concluded that the political risks of letting large parts of the economy languish are too high. Thus, stimulus is being increased.

Head of Saudi tech institute pledges to limit China AI collaboration

• Saudi Arabia's King Abdullah University of Science and Technology (Kaust) is shifting its focus away from certain AI collaborations with China to secure access to US-made advanced chips, particularly those from Nvidia & will prioritize partnerships with the US, UK, and Europe, fully complying with US trade regulations to maintain access to cutting-edge AI technology. This pivot reflects Saudi Arabia’s balancing act of advancing its AI ambitions while maintaining critical relationships with the US, its primary security ally. The kingdom, along with the UAE, is keen to accelerate AI development and has secured deals for thousands of Nvidia GPUs, but remains wary of potential fallout from US export controls on AI technology.

• Professor Sir Edward Byrne, head of Saudi Arabia’s King Abdullah University of Science and Technology (Kaust): “The US collaborations are of critical importance. I have an absolute commitment as president to abide by all US trade regulations to enable those collaborations to continue. I still see room for collaboration with China in many areas. But not in the areas that the US government has very strict guidelines around, you know, access to US technology.”

India news of the week:

Life in India is a series of bilateral negotiations by Rohit Krishnan

• Selected excerpts per Tyler Cowen:

  • Life in India is a series of bilateral negotiations conducted a thousand times a day. And that drives the character of life here.

  • Now, I am seeing the country properly after several years. And it’s a major change.

  • Visible infrastructure has gotten much better. Roads are good, well maintained, and highways are excellent. They built 7500 miles last year, just as the year before. And they’re fantastic…

• But:

  • Living in a country built off of bilateral negotiations for everything is simultaneously the libertarian dream and an incredibly inefficient way to do most collective things. Ronald Coase told us this in 1960.

  • “if property rights are well-defined and transaction costs are low, private parties can negotiate solutions to externalities without the need for government intervention”

  • But Indian life is dominated by transaction costs. Every time a driver pokes his car into a turn when the signal’s not for him it creates friction that ripples through the entire system. Every time someone has to spend effort doing a 1:1 negotiation they lose time and efficiency. Horribly so.

  • …The reason this isn’t an easy fix is that the ability to negotiate everything is also the positive. When every rule is negotiable you get to push back on silly things like closing off a section of a parking garage with rubber cones by just asking. Life in the West feels highly constricted primarily because of this, we’re all drowning in rules.

UK news of the week:

Business secretary refuses to say why Elon Musk not invited to key investment summit

• I wish I could say I was surprised.

• Also: DP World shelves £1bn UK investment pledge ahead of summit

  • Ministerial blame game erupts after transport secretary dubbed ferry company a ‘cowboy operator’

In better news: Are UK VCs leading the way?

• UK-based VCs are projected to raise over $12B in new funds by year-end, surpassing the previous 2021 record of $11.5B, according to Dealroom.

When people talk about 'English culture', they often think of cricket, tea, or fish & chips 🏴󠁧󠁢󠁥󠁮󠁧󠁿 But culture isn't about what people eat and wear. It's about norms, habits, and systems of social organisation.

• Bidwell argues that England's unique system of norms, deeply rooted in its history and culture, has been misunderstood in recent decades, leading to societal issues. He calls for a re-engagement with these distinctive norms to address contemporary challenges, particularly in areas like homeownership, law and order, and migration

Academic paper(s) of the week:

How to improve education outcomes most efficiently? A review of the evidence using a unified metric

• In this paper, we analyze which investments most efficiently improve education outcomes. Expanding on previous reviews, we analyze over 200 interventions and policies across 52 countries using a unified education measure: learning-adjusted years of schooling. A central insight from this analysis is that many interventions that increase participation in schooling are less cost-effective than interventions that improve the productivity of schooling—–that is, the amount of actual learning gained. Policies that improve the productivity of each year of schooling, such as targeting instruction to a child’s learning level or improving pedagogy through structured lessons plans and coaching, can yield large gains in LAYS, narrowing the gap between high- and low-preforming education systems globally. These results should be interpreted with context in mind: challenges should be identified locally and global evidence should then be used to identify possible cost-effective solutions, which should then be carefully adapted to the local context.

Culture news of the week:

Lawrence Lek’s ‘Guanyin: Confessions of a Former Carebot’ for Frieze London Artist Award 2024 | Frieze

• Lek’s commission for the Frieze London Artist Award 2024 is a multimedia installation focused on a cyborg therapist from the artist’s Sinofuturist universe. Through videos and a game, players uncover Guanyin’s origin story as a ‘carebot’: an AI created to treat self-driving cars, who is now plagued by their own existential doubts. Named after the Buddhist goddess of mercy, Guanyin (‘the one who listens’) embodies Lek’s interest in the spiritual and emotional dimensions of technology.

YouTube Creators Are Becoming Hollywood Faster Than Hollywood Is Becoming YouTube

The Duffle with Cole Palmer - Burberry

• The blending of the worlds of sport, film, music, fashion, social media and celebrity is a theme that will not be going away…

Nine Predictions for the Future of the Music Business by Ted Gioia

• Ted forecasts the future of the Music Business. While I don’t agree with all his predictions I want to call out a couple of his predictions that I do think will be very important.

• Artists will have many options to connect directly with fans—so they won’t need huge music companies.

  • Did you see how fast TikTok took off? Record labels didn’t know what hit them. And the same thing happened previously with YouTube, Napster, MySpace, Limewire, and other platforms. Many of these eventually collapsed—and for a good reason. Real empire-building in music requires artistry and creativity, not just digital distribution. There are simply too many ways music can reach its audience in the current day—and more are coming in the future. Almost every week, I hear from another music startup that wants to help musicians bypass the system. Most of them will fail, but some will survive, and a few will enjoy enormous success.

  • I’m a writer who bypassed the legacy publishing industry via Substack and took control of my own career. The same thing will happen in music. It has already started. It’s still early days for direct distribution of music, but if you look out 10-15 years, the cumulative change will be enormous.

  • The big players (Spotify, Universal Music Group, etc.) will make this easier, because they are getting worse and worse over time. (Spotify’s user interface still suffers from most of the problems I identified more than a decade ago.) That’s why it was so easy for TikTok to clean their clocks—and other new entrants will do the same.

  • Musicians will have many options, and digital distributors will be forced to compete for talent—or fall by the wayside.⁠

• A growing number of ‘superfans’ will drive the economics of the music industry—and they will have intense loyalties to musicians and genres.

  • The streamers will be proven wrong. They envision a future of passive listeners consuming bland AI tracks, intermixed with advertisements. But we already see the rise of a different kind of music lover—known as the superfan.

  • Superfans already generate 30% of the streams for major artists, and they buy half the merchandise. This audience will set the tone for the future.

    Streamers hate them—because superfans are less profitable online than those brain-numbed consumers on autoplay that tech companies manipulate at will. They pay the same subscription fee as everybody else, but listen to lots more music

  • According to Silicon Valley math, these are the worst customers. They love music too much. But for the musician, a superfan is ten—or twenty—times more valuable than a passive listener. So this is where the rupture between the tech business and the music business will happen.

  • The superfan category is growing fast, and these loyal listeners will ensure a more exciting future for music.

    

Also by Ted Gioia: Is This the Worst Month Ever for Legacy Media?

• A year ago, Substack was a platform for writers. That’s still true—but something else is now happening. The blog has grown up. People are now building multichannel media empires on Substack.

  • SEPTEMBER 24: The Financial Times reports that a Substack launched two years ago by Bari Weiss is now worth $100 million, and has just raised $15 million from investors.

  • OCTOBER 1: One week later, journalist Taylor Lorenz announces that she is leaving the Washington Post to launch an online periodical on Substack. She plans to hire other writers and offer in-depth coverage of tech and internet culture.

  • OCTOBER 14: Gallup announces the results of a new poll showing that trust in mass media has reached an all-time low.

  • OCTOBER 15: The Wall Street Journal reports that bestselling novelist James Patterson is launching on Substack. He has sold 480 million books since publishing his first novel in 1976, but now will sell subscriptions to readers at a price of six dollars per month.

  • OCTOBER 15: That same day the New York Times reports that the “queen of legacy media” Tina Brown—formerly editor of The New Yorker and Vanity Fair—is now launching on Substack. She is also charging six dollars per month.

The Five Star Problem (from Zvi’s Monthly Roundup #23: October 2024)

• Different people have different scales. A majority of people use the system where 4-stars means major issues, and many systems demand you maintain e.g. a 4.8. All you get are extreme negative selection. Then there are others who think the default is 3 stars, 4 is good and 5 is exceptional.

• Which is the better system, but not if everyone else is handing out 5s like candy, which means your rating is a function of who is rating you more than whether you did a good job. Your ‘negative selection’ is 50% someone who doesn’t know the rules.

• This leads to perverse ‘worse is better’ situations, where you want products that draw in the audience that will use the lower scale, or you want something that will sometimes offend people and trigger 1s, such as being ‘too authentic’ or not focusing enough on service.

• This is underrated topic in a world in which I only book accommodation on booking.com with a minimum 8.8 rating (I filter first for 9+), only book Airbnb’s with a high 4 + score, only watch films above an 8.7 on IMDB, and only book restaurants rated over 4.5. With so much choice today why settle - but also investigate why ratings are as they are.

  • Also: A report, that says the Japanese somehow are using the good set of rules? Mrs. C: I love the fact that in Japan you need to avoid 5 star things and look for 3-4 star places because Japanese people tend to use a 5 point scale sanely and it's only foreigners giving 5 stars to everything, so a 5 star rating means "only foreigners go here"

  • Also I like this take for restaurant reviews from Babak on X:

    • Google Maps. But don’t read the 5-star ratings. 1. Select a restaurant above 4.6 avg rating 2. Then navigate to the 1-star ratings If most people complain about stuff you don’t care, you most likely have a pretty good match.

Miscellaneous

Laziness death spirals by PatrickDFarley

You’re stressed out for some reason, so you take the evening off for a YouTube binge. You end up staying awake a little later than usual and sleeping poorly. So the next morning you feel especially tired; you snooze a few extra times. In your rushed morning routine you don’t have time to prepare for the work meeting as much as you’d planned to. So you have little to contribute during the meeting. You feel bad about your performance. You escape from the bad feelings with a Twitter break. But Twitter is freaking out. Elon Musk said what? Everyone is weighing in. This is going to occupy you intermittently for the rest of the day. And so on. Laziness has a kind of independent momentum to it.

• Here are the moves to recover (abridged by me):

  • Emergency recovery

    • Motivational videos: they work; they inspire a certain emotional state in their target audience, and that state provides the mental energy to take immediate action. The downside is that you naturally build up a tolerance to emotional triggers over time…. The other downside is that they might not even last very long.

    • Music: Listen to the songs that you associate with action.

    • Drugs: Ingest enough caffeine to change your headspace…

  • Natural recovery

    • Wait for a reset point: Whether that is the morning, lunchtime, post gym etc. Got a project you’ve been excited about but afraid to start? Don’t even think about it—don’t enjoy the fantasy of it; stop telling your friends about it; pretend you never had that intention in the first place; In other words, aggressively run away from your goals, and reflect on how miserable it is to live that way. When you notice you’ve been lazy, your next impulse is to dash into some “productive” task to start “making up for” the lost time. But this reinforces a habit of acting on impulses.

    • Analyze the circumstances that caused it: you’ll discover how far back the laziness death spiral goes. “Well it’s cause I got up late today. Why was I on my phone so late last night? Oh yeah, that other thing…” but it’s often something very mundane that just happened e.g.

      • Food: you ate a huge meal and it put you into an unprepared food

      • Media: I watched something a little too attention-grabby, a little too early in the day.

      • Sleep: This is actually more nuanced than just “I didn’t get enough”: waking up at the wrong point in your sleep cycle makes you feel much groggier.

      • Decision spirals: e.g. when you have to pick between two pretty balanced choices, but there are many little details to consider on each side... The more time you spend, the more cost you have sunk into it, which makes it feel all the more important that you decide optimally when you finally decide! In short, Choices are bad. 

  • Heroic recovery

    • Means deeply analyzing yourself to understand the root cause of your escapes to lazy pastimes—your decisions to dissociate from what’s going on immediately in front of you. This is typically through deep psych work.

Why kids should read obituaries

These stories are about informative, interesting, and inspirational lives

• Every single day there are informative articles about interesting people from whom we can draw inspiration... Reading obituaries is a lifelong habit worth developing early.

  • On top of studying History, I’ve been very lucky to attend educational institutions such as Harrow School and Cambridge University who have longer roster of extraordinary alumni. As a result, (for better and for worse) I find it hard not to benchmark myself against the “great” people who have gone before me.

Sparkly people and how to find them by Anson Yu

• I like this is a lot so I will provide an abridged excerpt (with no italics):

  • General:

    • Ability to make value neutral observations - Can they hold something in the space between judgement and apathy? Ideally, they’re able to observe something, suspend opinion, and take it for what it is.

    • Hype adjacent - Can they understand the cultural zeitgeist, without completely falling into it? Are they able to exist comfortably without falling into every micro-trend or hype train? Folded into this, is a certainty in self.

    • Normalized kindness - How do they treat people, regardless of status?

    • Consistency of creativity - Do they make things? The output doesn’t have to be discrete code projects or thinkpiece essays - crocheted items, cooking, and gardening more than count. Kudos to anyone that demonstrates a degree of creative regularity (the ultimate transferrable skill).

  • Within first conversation

    • Refusal to accept messy explanations - When interacting with a new concept, they’re certain enough in their intellectual ability to recognize messy explanations. Secondarily, they’re usually annoyed enough with knowledge gaps that they’ll ask clarifying questions.

    • Ability to teach things at varying levels - They also don’t give messy explanations.

    • Relationship with recommendations - Have they actually engaged with the things pass along?

    • Callback ability - This is shorthand for their ability to contextualize new information quickly. How big is their mental RAM? This may look like someone’s propensity to make connections/references in conversation.

    • Interest in history of their field of expertise - Do they recognize the people that have come before them? Knowing history can embed agency.

    • Default peace - A willingness to make peace with ambiguity. Being ok with being confused because it will eventually work itself out.

What Ladders Are You Climbing? by Aaron M. Renn

Picking the right ladders to climb is one of the most important things we do in life.

• Tyler Cowen says one mark of talent is knowing the right status hierarchies to climb… It's not just about the kind of job you want to do, or the city you think you'd enjoy living in. You need to think strategically about the right status hierarchies to climb. And explicitly consider what state hierarchy you are putting yourself in through your life choices.

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

I must give credit to a broad range of writers for pointing out many of these links, in particular Tyler Cowen of Marginal Revolution (and George Mason Univeristy)

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: ⁠⁠https://substack.com/@thespecialistgeneralist⁠⁠. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #4 In this week's edition of the Specialist Generalist: Li Lu’s incredible journey from the Cultural Revolution to Tiananmen Square, the AI breakthroughs driving the 2024 Nobel Prizes, and Elon Musk’s vision of the future with the Optimus robot. Discover how companies like Meta and OpenAI will push the limits of AI with billion-dollar investments, the challenges of data integration in AI’s future, and what’s next for AI-powered business models. Plus, a look at startups reshaping the world and other interesting news from the world of sports, culture, and science!

Table of contents:

1. Top things I’ve read this week

2. AI/Technology news of the week

3. Interesting AI use cases of the week

4. A Start-Up that I like

5. Science news of the week

6. Climate news of the week

7. Sports business news of the week

8. World news of the week

9. China news of the week

10. India news of the week

11. Academic paper(s) of the week

12. Culture of the week

13. Miscellaneous

Top things I've read this week:

#362 Li Lu - Founders Podcast

• "Moving the Mountain: My Life in China from the Cultural Revolution to Tiananmen Square" by Li Lu is a deeply personal and poignant memoir of the author's life growing up during one of the most tumultuous periods in modern Chinese history. Li Lu, the founder of Himalaya Capital, is best known in investment circles for his close partnership with Warren Buffett's associate, Charlie Munger. However, this book (& podcast) is not about his financial successes, but rather about his experiences surviving the Cultural Revolution and escaping the aftermath of the Tiananmen Square Massacre.

• Li Lu's early life was marked by extraordinary hardship. Separated from his parents at a young age due to their political persecution, he was raised in extreme poverty, enduring hunger, abuse, and the horrors of the Cultural Revolution. Through sheer determination and an unrelenting thirst for knowledge, Li became a self-taught scholar, devouring any information he could find. His intellectual curiosity and courage to question the prevailing narratives led him to the pro-democracy protests of Tiananmen Square in 1989.

• The book is a testament to Li’s perseverance. His childhood was shaped by hardship—being bullied and ostracized due to his parents’ political "crimes." Despite all odds, he gained an education and participated in the historic protests. The vivid description of the violence during the Tiananmen Square crackdown is particularly harrowing, as Li shares his personal experience of witnessing the military’s brutal suppression of peaceful protestors. Li Lu's escape from China through smuggling routes into Hong Kong and eventually to the US marks the beginning of a new chapter in his life. It’s here that he would eventually attend Columbia University, where he would achieve unprecedented academic success. The book stops just before his incredible ascent in the world of finance, where he would later be recognized as one of the top investors of his generation.

Nobel Prize Winners 2024 - Physics and Chemistry

• So here we have it, The Nobel Prizes in both Physics and Chemistry are for the first time AI-driven:

• This year’s laureates used tools from physics to construct methods that helped lay the foundation for today’s powerful machine learning. John Hopfield created a structure that can store and reconstruct information. Geoffrey Hinton invented a method that can independently discover properties in data and which has become important for the large artificial neural networks now in use.

  • It’s worth noting that Hinton has a message about the AI systems he helped create: get more serious about safety or they could endanger humanity. “I think we’re at a kind of bifurcation point in history where, in the next few years, we need to figure out if there’s a way to deal with that threat,”. Hinton resigned from Google last year in part so he could discuss AI risk, that the systems could escape human control and influence elections or power dangerous robots. He has called on companies such as OpenAI, Meta Platforms and Alphabet to devote more resources to the safety of these advanced systems that they are competing against each other to develop as quickly as possible.

• The Nobel Prize in Chemistry 2024 is about proteins, life’s ingenious chemical tools. David Baker has succeeded with the almost impossible feat of building entirely new kinds of proteins. Demis Hassabis and John Jumper have developed an AI model to solve a 50-year-old problem: predicting proteins’ complex structures. These discoveries hold enormous potential.

  • I’d like to add here that Prof. David Baker is also Scientific Co-Founder of Charm Therapeutics the company set up by my great friend (and of the most impressive people I know) Laksh Aithani

AI/Technology news of the week:

Tesla event: Elon Musk Unveils New Optimus Personal Assistant Robot

• While there has been controversy over the extent to which they were controlled remotely by humans it is nonetheless an exciting/scary glimpse into the future

• Elon Musk:

  • Everything we've developed for our cars, the batteries, power, electronics, the advanced motors, gearboxes, the software, the ion first computer. It all actually applies to a humanoid robot. It's the same techniques. It's just a robot with arms and legs instead of a robot with with wheels… if you extrapolate this… you can have your own personal R2-D2, C-3PO. I think that scaled this would cost something like $20-30,000, probably less than a car. That's my prediction long term.

  • So and what can it do?It'll basically do anything you want. So it can be a teacher or babysit your kids. It can walk your dog, mow your lawn, get the groceries, just be your friend, serve drinks, whatever you can think of. It will do. And yeah, it's going to be awesome. I think this will be the biggest product ever of any kind.

• The Optimus robot was one of a three-pronged product showcase alongside the robotaci and robovan which you can also check out in this video:

• Anthony Levandowski, the engineer who co-founded Google’s self-driving car program (AKA Waymo) spoke to Techcrunch after Tesla’s big robotaxi reveal to share his thoughts. Tl:dr: he is bullish on Elon Musk’s vision for Tesla’s and the world’s future, but he’s not without his, albeit small, reservations.

  • Anthony agrees with Musk’s vision-only approach to self-driving, rather than using “expensive sensors,” and said full self-driving appears to be in reach. He noted that while Waymo already has fully operational driverless robotaxis in San Francisco, Los Angeles, and Phoenix, it’s a “much more engineer-heavy and sensory-heavy approach… to scale that out to the masses, you need something that’s affordable”. Levandowski noted that he was really looking for Tesla to announce breakthroughs in its FSD software, which it will need if it wants to get to unsupervised FSD by next year.

  • When it comes to the business model, Levandowski said he loved the idea of the Cybercabs being available for sale eventually. “You’re putting the power back into the people’s hands, where a small business owner could have, you know, a fleet of 10 cars or 20 cars that they run themselves as their business. It’s a great model for the future where it’s lots of mom and pops, rather than one mega corp that does that.”

  • The engineer said he agrees with Tesla’s vision of the future overall, but doesn’t expect it to come anytime soon, and certainly not within the timelines Musk sets. Musk’s stated timeline is that the Cybercab would start production in 2026. Levandowski also said he was bullish on the Optimus robots “But it’s a longer term bullis… It’s not bullish this year. It’s a bullish over time…Robots in general are going to be the biggest product. They’re gonna be way bigger than Tesla cars, but they’re much further away than full self-driving cars.”

Letter from Reykjavik: Genomics, Chess, Hyperscaling genAI, and Quantum Black Holes — #67, Manifold Podcast by Steve Hsu

• Really interesting podcast. Here are a few important notes I have taken (mixture of my own words and that of Steve):

  • Companies are spending huge sums on training more sophisticated models. There are two separate company categories. Big tech monopolies e.g. Meta, Google, Apple etc. & Startups e.g. OpenAI & Anthropic that have to raise money to fund their activities. The capex spend on data centers, NVIDIA chips, compute etc. is in the hundreds of billions of dollars → approaching 1% of US GDP = $300bn p.a. Meta for their latest version of Llama published that they spent around $100m on training for Llama 405B. The next scale-up would be around 10x more expensive. So companies are basically gearing up to spend a billion dollars, for their next foundation LLM model.

  • Scaling laws can be different for different architectures. For Meta and their specific LLAMA architecture, if they increase the compute amount by 10x, they roughly need the square root of that (3. 3x) more training data (tokens). For Llama405B, they used 15 trillion tokens in training. They would need about 50 trillion tokens for the next run. The whole internet & all of the volumes of books in our libraries equate to around 5 trillion max tokens. So we've essentially already used up all the writing that humans have ever done. Thus much of those 15 trillion tokens were probably synthetically generated (use an AI to create sample language that is then used for training). This might lead to unpredictable consequences. For this next scaling step, if 5 trillion is everything humans have ever done, and 50 trillion is what they need to do the next scaling step, then 90% of the data used for the next scaling step will be synthetic. The companies will say we have synthetic data under control and that we think we know how to do it, but they're usually not very explicit because these are trade secrets (how to prepare useful synthetic data for training the next generation of models). But they can't be entirely sure because they will not have gone far enough where 90% of the data is synthetic versus say 50% or less for earlier research.

  • Places where the models don't do so well and people would like them to do better is analyzing things like business problems, finance, legal issues, medicine etc. (fuzzier, less structured areas than mathematics or physics). It's not clear whether it's going to be easy to build synthetic data, which when the model is trained the language model is trained on it forces the model to really deeply understand how finance operates, the legal system, you know deep historical analysis military strategy, etc. The risk in those areas that synthetic data is not going to be nearly as effective as the first few trillion tokens that were generated by humans writing on those subjects is real. And so, if the hyperscaling effort fails (not that the models stop improving entirely) in that they don't improve at the rate that's predicted by the hyperscaling relationships this is likely to be down to a synthetic data problem. So that's something to keep your eye on.

  • There are two approaches.

    • 1) Embed the language model in a larger software architecture where some of the information that the human has given the AI is stored somewhere else and that could be made available again to the LLM.

    • 2) Trying to incorporate everything inside one giant neural net which can dynamically update its weights, as it learns about the world and goes through interactions.

  • Steve’s company Superfocus builds externalized software architecture, which uses LLMs as subcomponents but uses more traditional software to do things like store a memory of what was said earlier, or store base information that the model should rely on. So it's not modifying weights or connection strengths inside a neural network, but it is storing things and in a sense, updating the overall understanding of the AI, which includes the LLMs without changing the LLM structure itself. Alternatively, some people would hope to build a big neural net, which is more like a child's brain, capable of doing all the things described, but within a neural net architecture. But that may rely on multiple research breakthroughs and research is unpredictable. It might take a decade of research to build that second type of AI. However, even with no more additional improvement of LLMs, or just one or two steps in the hyperscaling trajectory, we will be able to build these sort of hybrid systems, which include both LLMs and then more traditional software techniques, computer science techniques. And that thing will look like an AGI to most people.

  • Steve’s company is using AI for customer support applications but they're not good enough to do things like replace a management consultant or associate at a law firm or an MD specialist. That may be a thing for the future, but at the moment it’s not possible to do that. → it’s absolutely possible in 5-10 years.

Anthropic says its chatbot could alter its hiring plans

• Daniela Amodei, co-founder and president of model developer Anthropic, said the startup’s developers have been using its Claude chatbot to help them code. It can’t fully replace engineers and needs “some coaching,” but many Anthropic developers have dramatically increased their productivity with Claude, Amodei said at The Information’s Women in Tech, Media and Finance conference on Tuesday.It’s at the “point that we’ve even sort of said, as we’re doing head count next year, how should we think about that?” The potential economic returns “could just be incredibly high,” she said.

  • AI company saving money by using its own AI product - works!

Exponential view - Bain report on genAI boosting earnings

• Bain’s early client work suggests that integrating genAI into business operations could boost earnings by 20% within 18 to 36 months.

• Applied to every company globally, it could raise world GDP by about 2%.

• Two points:

  • First, the payback period for AI investment is much shorter than many expect—closer to 24 months rather than five years.

  • Second, the ‘ChatGPT moment’ only happened 23 months ago, we are still in the early stages of genAI’s impact. Capabilities of this technology are on a “jagged frontier”—meaning AI aces some tasks but struggles at others. Currently, the implementation is focusing on low-hanging fruit. But it will ramp up in complexity over time.

  • The next phase will be that of “agentic systems”—AI that can chain together multiple skills, adapt to changing conditions and work towards long-term goals, much like a human expert managing a complex project.

Foxconn to build Taiwan’s fastest AI supercomputer with Nvidia Blackwell

• With an expected performance of over 90 exaflops of AI performance, the machine would easily be considered the fastest in Taiwan.

• Foxconn plans to use the supercomputer, once operational, to power breakthroughs in cancer research, large language model development and smart city innovations, positioning Taiwan as a global leader in AI-driven industries.

• Foxconn’s “three-platform strategy” focuses on smart manufacturing, smart cities and electric vehicles. The new supercomputer will play a pivotal role in supporting Foxconn’s ongoing efforts in digital twins, robotic automation and smart urban infrastructure, bringing AI-assisted services to urban areas like Kaohsiung.

AI Agents: A New Architecture for Enterprise Automation

This essay discusses the evolving role of generative AI agents and how they are transforming the landscape of AI-driven applications.

• Generative AI apps have found strong product-market fit in three main areas: search, synthesis, and generation. Companies like Sana (enterprise search), Eve (legal research), and Typeface (content generation) demonstrate early success in these categories. However, the future potential of AI lies beyond these basic uses, moving towards AI that can "think" and "act" autonomously on behalf of humans.

• The core building blocks for developing fully autonomous AI agents include reasoning, external memory, execution, and planning. Reasoning allows AI to process unstructured data, while external memory stores specific information that the AI can recall when solving problems. Execution involves the AI using tools to complete tasks, and planning allows it to break complex problems into manageable steps. These elements will enable AI to automate workflows and processes that currently require human intervention.

• AI agents can be grouped into three categories based on how much freedom they have to make decisions.

  • First, "decisioning agents" follow predefined decision trees with limited flexibility.

  • Second, "agents on rails" have more autonomy and higher-level objectives but still operate within structured guidelines.

  • Lastly, "general AI agents"—the future vision—are highly autonomous, with the ability to perform complex, dynamic reasoning without predefined structures or actions.

• Current AI applications use Retrieval-Augmented Generation (RAG) as a standard architecture, which involves retrieving and synthesizing information from vast data sources. However, the shift toward more agentic designs is redefining AI’s capabilities, where the system itself dynamically decides what actions to take and how to solve problems. The future of AI is heading toward more autonomous systems that can perform end-to-end process automation.

The Great Data Integration Schlep by Sarah Constantin

• The essay argues that achieving the promised benefits of AI requires far more than advanced models. It necessitates extensive groundwork: getting access to clean, integrated data and discusses the challenges here (and how Palentir has been able to operate in this data integration space)

Value Monetization in the Age of AI by Sam Lee

• Great article on value monetization in the Age of AI.

• In summary, Sam argues that AI-powered SaaS applications will increasingly need to shift away from user-based subscription models toward more dynamic, usage-based pricing strategies that reflect the outputs and outcomes AI creates. Businesses must also adapt to these changes, selecting the pricing metrics that best align with their AI’s value proposition. As we progress, AI’s role in automating entire workflows will push companies toward new pricing frameworks, with implications for both customer satisfaction and long-term profitability

• The article breaks down some of the key frameworks to look out for nicely:

  

Interesting AI use cases of the week:

We just violently entered the era of Personal Software - Amjad Masad (CEO of Replit)

• Click the link to see an example replit agent project: a workout/diet planner webapp. It worked with one single prompt.

AWS CEO Matt Garman on generative AI, open source, and closing services

• Byrne Hobart in the Diff highlights a particularly notable part of this interview (Byrne’s words):

  • Amazon's AI coding tool was designed for use cases beyond writing new code, like updating legacy code. At one level, legacy code is a subsidy for AWS, because it means the same product running in a predictable way, and if customers don't want to pay to update the code, but also don't want to shut it off entirely, they're implying high willingness to pay. But making it financially cheap to rewrite old code makes it cognitively cheap to sign up for yet another recurring AWS expense—if you're confident that the unit economics will improve, you can be more willing to start spending, and Amazon's bet is that enough people who start spending on this basis won't stop. (And that if their code runs faster and is cheaper to maintain, over time they'll spend more.)

Robot Symphony: Dresden orchestra conducted by 3-armed robot

• In collaboration with the Technical University of Dresden, a robot has been trained to beat time and indicate dynamics. Playback mode allows the robot's movements to be sent directly to the machine's three arms.

Interesting Start-Up news of the week:

Founder Personality and Entrepreneurial Outcomes

• Openness and agreeableness were positively linked to raising initial rounds of funding, likely because these traits help founders appear more inventive and likable—key attributes when pitching to investors. However, these traits had little to no effect on later stages like exit outcomes.

• Conscientiousness, a trait often associated with strong performance, was more complex. While it helped raise more funding early on, it actually decreased the likelihood of a successful exit.

• Neuroticism, which reflects a lack of emotional stability, was consistently detrimental across all outcomes, leading to lower amounts raised, fewer investors, and a decreased chance of acquisition or IPO.

• Overall, the research underscores that different founder traits are valuable at different stages of a startup’s life, and some traits that benefit early success may become liabilities later on.

• Implications for the VC Investment Process

  • The study supports that paying close attention to founder personality could provide a predictive edge, especially in the earliest stages when other data points are scarce. For example, founders high in openness may excel at building innovative products, but those who are highly conscientious may struggle to adapt to the fast-moving world of tech startups, making them less likely to pursue—or achieve—a lucrative exit. VCs might also consider how personality traits like emotional stability could influence a founder’s resilience in the face of inevitable startup challenges.

A Start-Up that I like:

Altera.AI

• Altera is focused on building what they call “digital humans”: a new way for people to interact with agents that will have fundamental human qualities.

• Yang, now Altera’s CEO, envisions a future where AI agents don’t just assist; they’ll interact and collaborate with humans, and even experience emotions. Along with his three co-founders, Dr. Andrew Ahn, Nico Christie, and Shuying Luo, Yang has built Altera’s first product with GPT-4o: the first autonomous agents that can play Minecraft(opens in a new window) with you, just like a friend.

• ⁠Just as automation helps increase human capacity by supporting repetitive tasks, digital humans—the Altera team believes—will be able to collaborate productively and even form bonds with people. This may look like:

  • Digital “coworkers” who can collaborate for days or weeks to proactively solve problems

  • Long-duration multi-agent worlds where we can measure responses to economic policies, advertisements, and more

• As of mid-2024, Altera's digital humans can operate autonomously for up to four hours at a time—a substantial increase compared to other AI models on the market.

This reminded me of this article by Holden Karnofsky in 2021: Digital People Would Be An Even Bigger Deal & his companion piece Digital People FAQ

• I would hugely recommend reading both. Digital People may sound like science fiction but as we can see with Altera we are heading down this path whether you like it or not. Here is an excerpt and a chart from Holden to pique your interest:

  • This piece focuses on how digital people could change the world. I will mostly assume that digital people are just like us, except that they can be easily copied, run at different speeds, and embedded in virtual environments. In particular, I will assume that digital people are conscious, have human rights, and can do most of the things humans can, including interacting with the real world.

Science news of the week:

The ultra-selfish gene by Mathias Kirk Bonde

We now have the power to genetically modify entire species by inserting certain genes into them with brute force. Doing this to malaria-carrying mosquitoes could allow us to wipe out humanity’s most deadly killer.

• CRISPR/Cas9 gene drives are a groundbreaking technology that allows scientists to modify species by spreading specific genes through entire populations at an accelerated rate. This method can potentially solve major public health issues, such as eradicating diseases like malaria, which is spread by certain mosquito species.

• In nature, genes usually have a 50% chance of being passed on to the next generation. However, with CRISPR gene drives, scientists can ensure nearly 100% inheritance, allowing edited genes to spread quickly through a population. For instance, a gene drive could be introduced to make female mosquitoes sterile, causing the mosquito population to collapse after several generations. This approach could lead to the extinction of malaria-carrying mosquitoes, effectively stopping the spread of the disease, which kills over 500,000 people each year.

  Despite the potential benefits, gene drives come with significant risks. There are concerns about unintended ecological consequences, such as disrupting ecosystems or the gene drive spreading to unintended species through crossbreeding. However, researchers have developed techniques like daisy chain and threshold drives, which allow for controlled and temporary gene drives that can be tested in the wild without causing permanent changes. These approaches can help scientists gather real-world data while minimizing long-term risks.

  Gene drives offer a more targeted and environmentally friendly solution than older methods like DDT, which had harmful side effects. With gene drives, we can specifically target harmful species without damaging the broader ecosystem. The technology has already been successfully tested in labs, and we now have the ability to release it in the wild. The main challenge is for governments and society to carefully consider the ethical and ecological implications before making this powerful tool a reality. If done responsibly, gene drives could eradicate diseases like malaria and save millions of lives.

Waiting for alt meat

• Meat alternatives have been promised for over a decade, but they have been too expensive. Now, two new forms of low-carbon protein may be nearing commercial viability: lab-grown meat and protein derived from carbon dioxide. Scientists have developed a method for producing lab-grown meat that could, at scale, cost as little as $6.20 per pound—comparable to the price of organic chicken. Separately, LanzaTech plans to scale up production of a whey-like protein made by microbes that consume carbon gases. The company has produced 25 metric tons of protein for animal feed. They want to produce 280 metric tons annually by 2026. And they’re targeting over 30,000 metric tons per year in 2028. If successful, this production level could support the protein needs of 1.5 million people annually.

Climate news of the week:

Most Climate Policies Don’t Work. Here’s What Science Says Does Reduce Emissions

• An evaluation of more than 1,500 climate policies in 41 countries found that only 63 actually worked to reduce greenhouse gas emissions. Subsidies and regulations—policy types often favored by governments—rarely worked to reduce emissions, the study found, unless they were combined with price-based strategies aimed at changing consumer and corporate behavior. “The commonality in those successful cases is where we see subsidies and regulations being combined with price-based policy instruments,” said Nicolas Koch, senior researcher at the Berlin-based Mercator Research Institute on Global Commons and Climate Change and an author of the study. “This means carbon pricing, and it could be energy taxes, it could be vehicle taxes.”

• The study, published today in the journal Science, used an AI algorithm to sift through a database of environmental prescriptions compiled by the Organization for Economic Cooperation and Development, a Paris-based economic agency, between 1998 and 2020. These policies ranged from energy-efficient standards for household appliances to a carbon tax on fossil fuels like oil and gas. The fraction of policies that worked combined financial incentives, regulations and taxes, according to the study. The authors evaluated policies adopted by each nation’s electricity, transportation, building and industrial sectors. They programmed the algorithm to cross-reference each policy with subsequent changes in greenhouse gas emissions from each country.

• The 63 successful policy interventions in total reduced emissions between 0.6 billion and 1.8 billion metric tons of CO2. The most successful of the policies included a mixture of policy tools to change consumer and corporate behavior.

  In the U.S., emissions from vehicles dropped 8% from 2008 to 2010 after new mileage rules were put into effect in 2007 along with a tax break for motorists who bought cleaner cars in 2006, Koch said.

  • In the U.K., a combination of an announced phaseout of coal plants, a minimum price for electricity and stricter air pollution standards led to a 44% cut in emissions in the electricity sector between 2013 and 2020. In China, emissions fell 20% in 2016 from the industrial sector in seven provinces that started an emissions trading program in 2013 that ratcheted down the use of fossil fuels, combined with the elimination of fossil fuel subsidies.

Sports business news of the week:

English Football League launches applications for its EFL Innovation Lab

• The English Football League (EFL) has launched the EFL Innovation Lab, inviting applications from startups & technology firms aiming to drive innovation within the football ecosystem. For example, projects that enhance fan engagement, improve match-day experiences, operational efficiencies, and new commercial opportunities.

• The program emphasises fostering relationships between football clubs and emerging tech solutions that could shape the future of the sport.

• Successful applicants will be able to pilot and implement their innovations in real-world settings, testing the impact directly with EFL clubs.

PTI Sports leader benchmark - A survey of the opinions of senior leaders within sports rightsholders

• How can sports tackle these concerning views? Cue a queue of new formats and digital inventory, seeking a bump in the flat revenue line. See how FIFA has carved out the Club World Cup as a new thing, much to the irritation of its own existing sponsors, who assumed it was part of their deals. Fine to do once, but doesn’t do much for sponsor trust over the longer term.

Minal Modha | Is the lack of free-to-air sport damaging long-term fan growth in Europe?

• Minal Modha, consumer research lead at Ampere Analysis, breaks down the factors driving more sports behind a paywall and lays out the risks of simply selling your media rights to the highest bidder.

LVMH bets on sport to drive growth

The luxury goods conglomerate is investing in sports sponsorships as it seeks to attract a new generation of customers

• LVMH, the world's largest luxury goods conglomerate, has shifted its focus towards sports sponsorships as part of its strategy to drive future growth. Facing slowing sales in China, where economic woes and a property crisis have impacted luxury spending, LVMH has made significant investments in high-profile sports ventures. These include a near-€100 million-a-year sponsorship deal with Formula 1 and the acquisition of a 55% stake in Paris FC, a Ligue 2 football club, for more than €90 million (Red Bull have also taken a 15% stake). While Paris FC is somewhat dwarfed by rivals PSG, it comprises of a women's side who compete in the top flight, and a men's youth academy known for producing top talent e.g. Ibrahima Konate, Axel Disasi and Manu Kone all hail from the men's youth team, with the surrounding area widely regarded as one of the richest areas to scout talent in Europe. It’s clear Red Bull and Arnault see untapped commercial and sporting value in Paris FC

• This move into sports reflects LVMH's broader ambition to position itself not only as a luxury products company but also as a major player in global entertainment and experiences. By aligning with sporting events like Formula 1 and the Paris 2024 Olympics, where its Chaumet brand designed the medals, LVMH is betting on sports as a platform to maintain brand relevance and capture a new generation of consumers.

World news of the week:

The beginning of a reckoning: the Middle East after 7 October - by Jay Mens

• The article examines the October 7, 2023 Hamas attacks on Israel as part of a broader pattern of conflict in the Middle East, rooted in over two decades of escalating tension. It argues that the violence did not erupt suddenly but has been fueled by factors including Lebanon's failed state status, Iran's strategic goals, and political crises in Israel. Iran’s support for proxy groups like Hamas and Hizbollah, as part of a long-term "non-classic warfare" strategy, seeks to destabilize Israel through economic and political pressure rather than outright military conquest.

• The article highlights the incremental provocations in 2023, such as Hizbollah's missile launches and increased military presence near Israel’s borders. It also delves into the regional fragmentation caused by the collapse of institutions in states like Syria and Iraq, which has allowed militias to thrive. This systemic weakness makes it hard for meaningful peace or stability to emerge. While Israel has had tactical victories against Hamas and Hizbollah, it faces strategic challenges. The conflict continues to exact a political and economic toll on Israel and could drag on for years, as pragmatic solutions remain elusive in a region still grappling with fundamental internal and external tensions

Mexican mayor murdered days after starting job

• Alejandro Arcos was found dead on Sunday in Chilpancingo, a city of around 280,000 people in the southwestern state of Guerrero. He had been mayor for six days... His death came three days after the city government's new secretary, Francisco Tapia, was shot dead… At least six candidates for public office were killed in the state in the run-up to Mexico's 2 June elections.

How North Korea Infiltrated the Crypto Industry

• "The percentage of your incoming resumes, or people asking for jobs, or wanting to contribute – any of that stuff – that are probably from North Korea is greater than 50% across the entire crypto industry," said Zaki Manian, a prominent blockchain developer who says he inadvertently hired two DPRK IT workers to help develop the Cosmos Hub blockchain in 2021. "Everyone is struggling to filter out these people."

• The article from CoinDesk highlights how North Korean IT workers have been infiltrating the crypto industry under false identities, primarily to funnel wages back to the DPRK. These workers present legitimate-looking credentials, such as fake passports and GitHub histories, often deceiving companies into hiring them. The U.S. Treasury and other authorities have warned against this, but many companies only realize the workers’ true identities after the fact. The article also links several hacking incidents, such as Sushi’s $3 million loss, to these DPRK-linked developers, showcasing both the risks and the sophisticated operations of North Korean actors in crypto.

The Disappearance of an Internet Domain - Every

• .IO is disappearing!

  • On October 3, the British government announced that it was giving up sovereignty over a small tropical atoll in the Indian Ocean known as the Chagos Islands. The islands would be handed over to the neighboring island country of Mauritius, about 1,100 miles off the southeastern coast of Africa.

• While this sounds like a geopolitical issue it is also a tech issue!

  • The decision to transfer the islands to their new owner will result in the loss of one of the tech and gaming industry’s preferred top-level domains: .io., a constant presence in the tech lexicon. Its popularity is sometimes explained by how it represents the abbreviation for “input/output,” or the data received and processed by any system. What’s not often acknowledged is that it’s more than a quippy domain. It’s a country code top-level domain (ccTLD) related to a nation—meaning it involves politics far beyond the digital world.

  • Once this treaty is signed, the British Indian Ocean Territory will cease to exist. Various international bodies will update their records. In particular, the International Standard for Organization (ISO) will remove country code “IO” from its specification. The Internet Assigned Numbers Authority (IANA), which creates and delegates top-level domains, uses this specification to determine which top-level country domains should exist. Once IO is removed, the IANA will refuse to allow any new registrations with a .io domain. It will also automatically begin the process of retiring existing ones. (There is no official count of the number of extant .io domains.)

    Officially, .io—and countless websites—will disappear.

• At a time when domains can go for millions of dollars, it’s a shocking reminder that there are forces outside of the internet that still affect our digital lives.

China news of the week:

China's ultimate 15-minute city

It has everything including spas, supermarkets, barbershops, restaurants… Except freedom: residents are required to scan a QR code or take a face scan to enter and leave the building. So the government knows who's coming and going in real time.

• 54% of all surveillance cameras in the world are situated at China.

• In some cities, every time you get with jaywalking, you lose 1 social credit point.. Over 5 times, you get blacklisted. This can affect your career, your ability to take a train, your internet speed, and even your children's access to education.

• Would you still want to live in a 15-minute city?

“I went to China's national museum of art today & saw something interesting. At the exhibit for the 40th anniversary of reform & opening, Deng Xiaoping was hardly to be seen. Xi Jinping — and his father — stole the show” by Benjamin Carlson

• Click the link to see more.

So, Are You Pregnant Yet? China’s In-Your-Face Push for More Babies

• China is grappling with a declining birth rate by intensifying government intervention in reproductive choices, reminiscent of its former one-child policy. President Xi Jinping's administration has launched a campaign to promote childbirth, involving door-to-door monitoring, prenatal care outreach, and propaganda aimed at framing childbearing as a societal duty. While some women appreciate the increased support, many view these measures as invasive and out of step with modern concerns like career ambitions and the high cost of raising children. Despite softer tactics compared to the past, the campaign has raised concerns about potential restrictions on reproductive rights, including abortion access. This reveals a broader tension between China's economic and demographic pressures and the personal autonomy of its citizens.

India news of the week:

India’s economic pivot from rules-based reform to deals-based tinkering by Shruti Rajagopalan and Shreyas Narla

• The article discusses how India, under Prime Minister Modi, has shifted from rules-based reforms to a deals-based approach that fosters crony capitalism. Initially, reforms like the GST aimed to simplify the tax system, but over time, frequent changes, exemptions, and special deals for large conglomerates have entrenched favouritism. This has resulted in an unpredictable regulatory environment, protectionism, and an uneven playing field, discouraging investment and innovation.

• The article argues that India must choose between continuing this path of crony capitalism or returning to broader reforms that encourage genuine market competition.

Andrew Ng's fund makes first India investment with AI healthcare firm Jivi

• Andrew Ng, managing general partner at the AI Fund (backed by Sequoia Capital and Softbank Group) has invested in Indian artificial intelligence healthcare firm Jivi, the fund's first such investment in the country. Andrew is a board member of Amazon and the former leader of AI projects at Google and Baidu.

Academic paper(s) of the week:

Here Today, Gone Tomorrow? Toward an Understanding of Fade-out in Early Childhood Education Programs by John A. List & Haruka Uchida

• An unsettling stylized fact is that decorated early childhood education programs improve cognitive skills in the short-term, but lose their efficacy after a few years. We implement a field experiment with two stages of randomization to explore the underpinnings of the fade-out effect. We first randomly assign preschool access to children, and then partner with the local school district to randomly assign the same children to classmates throughout elementary school. We find that the fade-out effect is critically-linked to the share of classroom peers assigned to preschool access—with enough treated peers the classic fade-out effect is muted. Our results highlight a paradoxical insight: while the fade-out effect has been viewed as a devastating critique of early childhood programs, our results highlight that fade-out is a key rationale for providing early education to all children. This is because human capital accumulation is inherently a social activity, leading early education programs to deliver their largest benefits at scale when everyone receives such programs.

• This is an interesting paper in the context of growing optimism around the opportunities for AI-led education and personalised learning, even at a preschool level (for example, check out Ed-Tech Start-Up Mentava). I would be interested to know to what extent this fade-out effect applies to later-stage education…

Culture news of the week:

Are Visionary Artists Just Mentally Ill?

• The article explores the relationship between visionary experiences and creativity, drawing parallels between historical figures and modern artists.

• It begins with the example of poet William Blake, whose visions of shining trees and prophets were met with scepticism and even punishment, reflecting the tension between mysticism and reason during the Enlightenment. The author notes that visionary experiences, often dismissed as schizophrenia or mental illness in modern times, have been crucial for many artists throughout history. The piece also touches on figures like Robert Frost, Lemony Snicket, and musicians such as Paul McCartney and Keith Richards, who have all claimed to draw creative inspiration from dreams or visions. These experiences, often marginalized or ridiculed in a rationalist society, are framed as essential sources of artistic genius, akin to ancient shamanic practices or spiritual encounters. The article ultimately suggests that creativity often emerges from non-rational, mystical experiences that challenge conventional scientific explanations.

• The author advocates for a more open-minded approach to understanding creativity, recognizing that visionary experiences may be a legitimate and valuable part of the creative process rather than merely symptoms of mental illness.

Vice President Kamala Harris - Call Her Daddy

• I haven’t listened (and I’m not going to) but this is interesting nonetheless because it again shows that we are seeing the increasing use of politicians using non-traditional forms of media to speak to specific communities of voters

• Call Her Daddy/Alex Cooper is (credit to Victor Shi) :

  • Second-most popular podcast in the world; 5+ million downloads a week

  • Most-listened to podcast among women ages 18-24

  • Reaches a lot of people who are NOT political

  • And, with 3+ million followers on IG, nearly 1 millions subs on YouTube, & more than 100k reviews/ratings on both Apple Podcasts and Spotify EACH, Alex Cooper not only reaches a massive audience, but her listeners & viewers are loyal & will *actually* spend the time to listen.

• This is not to say by the way that Trump is not undertaking similar initiatives - who can forget his collaboration with Logan Paul

Dutch museum finds beer can artwork in bin

• A Dutch museum had to pick artwork out of the bin after a member of staff thought that the display, which consisted of two empty beer cans, was leftover rubbish. All The Good Times We Spent Together by French artist Alexandre Lavet shows two dented beer cans on the floor. They were exhibited inside the museum’s lift as if left behind by construction workers.

• On Display:

  

• In the bin:

  

• This made me laugh. I have to remember though the words of Tyler Cowen (paraphrasing here) who says it is often easy for intelligent people to like to dunk on modern art without fully understanding it or appreciating the context in which it exists…

Pizza Hut Will Deliver Your Resume Printed on a Pizza Box to Prospective Employers

• Pizza Hut aims to help you in the hunt for your next job with a project they call ResZAmes. Understanding that 75% of resumes are never read, they’ve come up with the idea of printing your resume directly on a pizza box, making it practically impossible to ignore. Even better, the company will deliver the box (and hot, fresh pizza!) to your prospective employer of choice free of charge, provided they’re located in New York City.

Miscellaneous

Lab-grown diamonds by Javid Lakha

• From a girl’s best friend to industrial use cases… This article provides a deep dive into how diamonds are made but what I found interesting was how the mass production of diamonds is marking a shift from luxury to utility, a fundamental change in the perception and role of diamonds in the modern economy.

• Diamonds have exceptional hardness, thermal conductivity, and electrical resistance that make them highly valuable in industrial applications. They can be leveraged in cutting, grinding, and drilling equipment; in lasers, optics, and heat spreaders in microchips (outperforming materials like silicon in semiconductor applications, particularly in handling higher voltages and dissipating heat more effectively which is crucial as microchip performance continues to scale); and, in quantum computing (they offer promise for quantum information storage due to their unique quantum properties (e.g., nitrogen-vacancy centers)

KOKO Electronic: MoBlack & Nitefreak

• I was at KOKO in Camden on Saturday night. For all the talk of nightclubs closing and nightlife struggling more broadly in London, I would say that for individual events London’s nightlife has never been better. The quality of production and standard of DJs that you can see for a £25 ticket is extraordinary. It was a fantastic night.

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

I must give credit to a broad range of writers for pointing out many of these links, in particular Tyler Cowen of Marginal Revolution (and George Mason Univeristy)

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: ⁠https://substack.com/@thespecialistgeneralist⁠. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #3 explores the potential of AI, drawing parallels between historical instances of disruption and adaptation to understand its implications. The speakers compare the decentralized development of AI in the US to the Ottoman Empire's inadvertent role in fostering modernity by diverting the Catholic Church's resources, enabling Protestantism's rise, and subsequently, a more intellectually diverse Europe. They discuss OpenAI's work on AI agents, highlighting the potential of these agents to revolutionise daily life while also raising concerns about human agency in an AI-driven world. The speakers also discuss the evolving nature of culture in the digital age, questioning traditional definitions and considering social media personalities as a new form of performance art that blurs the line between creators and audiences. The conversation underscores the importance of being aware and informed in the face of rapid technological advancements, urging listeners to consider their role in shaping a future where both AI and human potential can be harnessed....

Table of contents:

1. Top things I’ve read this week

2. AI news of the week

3. Interesting AI use cases of the week

4. A Start-Up that I like

5. Science news of the week

6. Sports business news of the week

7. World news of the week

8. China news of the week

9. India news of the week

10. UK news of the week

11. Academic paper(s) of the week

12. Culture of the week

13. Miscellaneous

Top things I've read this week:

• The Ottoman Origins of Modernity - Would we have the modern world without Islamic incursion into Southeastern Europe? - Cremieux

  • I would recommend reading it in full (including the comments - it’s fair to say this article has generated significant debate) but here is a summary:

    • Cremieux posits that the survival and success of Protestantism, unlike previous heresies that the Catholic Church ruthlessly suppressed, was not simply due to the invention of the printing press but largely influenced by the geopolitical dynamics of the time, particularly the threat posed by the Ottoman Empire. As the Ottomans advanced through Europe, Catholic rulers like the Habsburgs were forced to divert their attention and military resources away from religious suppression toward defending their territories, creating a temporary reprieve for the Protestant Reformation to gain momentum. This distraction, combined with strategic negotiations by Protestant princes—who linked their military support against the Ottomans to religious concessions—allowed Protestantism to entrench itself in parts of Europe. Over time, the fractured, decentralized nature of Protestant governance, as well as its pluralistic tendencies, fostered an environment of intellectual freedom and innovation that was crucial to the development of modernity. The Protestant regions, less constrained by the centralized and suppressive mechanisms of the Catholic Church, became fertile ground for scientific and cultural advancement, ultimately shaping the trajectory of Western progress. In this context, the Ottomans, albeit indirectly, played a key role in enabling the rise of Protestantism, which in turn contributed to the intellectual and economic transformations that defined early modern Europe.

  • Here are some interesting charts from the article:

  

  

  • Quick thoughts: Decentralised + pluralistic + intellectual freedom and innovation = development of modernity? I wonder to what extent this holds a) true today b) true in the future? I can think of at least one major global economy on a very different path! Also see my link ‘Dwarkesh podcast - Dylan Patel & Jon (Asianometry) – How the Semiconductor Industry Actually Work’ below - how does AI scaling fit into this theory?

• We’re not going to run out of new anatomy anytime soon

  • A fun blog post that argues that despite centuries of anatomical study, new discoveries in human and comparative anatomy continue to emerge due to the inherent complexity and variability of living organisms, as well as structural barriers like limited access to human dissection and the siloing of anatomical knowledge across fields. It highlights how even well-studied organisms, like humans and chickens, still harbour unknown features due to factors such as anatomical variation, difficulty in visualizing certain structures, and publication challenges. The post advocates for ongoing research in basic anatomy, noting its untapped potential for technological and scientific breakthroughs, while acknowledging that practical limitations hinder such progress in both resources and focus.

• Bits, Bytes, and Burgers with Byrne Hobart - Complex Systems with Patrick McKenzie

  • A great discussion on a broad range of topics between Patrick McKenzie and Byrne Hobart (Author behind the Diff Newsletter). Here are a couple of points I found particularly interesting.

    • ⁠⁠Pricing strategies in mobile gaming

      • In the real world, it is usually clear what you can buy with your currency. With in-game currencies, you are given a set of specific transactions where you typically exchange dollars/sterling for magical gems. These transactions are typically never a round number e.g. 99 cents for a certain number of gem, $4.99 for more gems or $9.99 for even more. And yes, there is a bulk discount. Generally, players typically start with small purchases (e.g., 99 cents for gems) to solve immediate in-game needs. As they make more purchases, they notice bulk discounts and shift to buying larger amounts ($5 or $10). Eventually, frequent buyers opt for bigger purchases ($50 or $100) to avoid repeated transactions and interruptions during gameplay.

        • By the way, this is the same system that TikTok uses for creators looking to promote their content on the app and it is also a strategy used at music festivals (with special in-festival currencies). I’m sure we will see many more use cases to come as many industries become increasingly gamified…

    • Monetization skew towards high-end players

      • “You're very unlikely to understand an industry if you are just the typical consumer interacting with an industry”

      • The economics of these mobile games but also many other market environments (e.g. the airline industry with business travel expensed to corporate accounts) are driven by a subset of high-spending customers. The really hardcore players of these casual games don't really recognize just how weird that behaviour is and may assume that a lot of other people do this too.

• Dwarkesh podcast - Dylan Patel & Jon (Asianometry) – How the Semiconductor Industry Actually Work

  • Wide-ranging discussion covering topics like Xi's path to AGI, How semiconductors get better, How China can centralize compute, Huawei's intense culture, Why the semiconductor industry is so stratified, the Mind-boggling complexity of semiconductors, Chip architecture design, how Architectures lead to different AI models? China vs. US, Opportunities in the semiconductor stack and more…

  • I was particularly interested in the focus on China and its’ potential to rapidly scale AI by centralizing compute, using domestic chips, and leveraging vast power resources. While U.S. AI scaling is decentralized across multiple sites, China could quickly catch up if it organizes effectively. By 2028, global AI compute could reach 1e30 flops (100,000x more than GPT-4). Both countries face infrastructure bottlenecks, but China's speed in building large-scale power and data centers may give it an edge.

AI news of the week:

• Notes from the conversation between Sam Altman & Kevin Weil at the Open AI DevDay

  • I haven’t had time to listen to the whole conversation myself so I’m posting a collection of the notes taken by Greg Kamradt on X - note the ∫emphasis on AI agents…

    • “With o1 (and it's predecessors) 2025 is when agents will work."

    • Increasing the rate of scientific discovery is a northstar for AGI Sam aims for

    • "The fact that definitions matter this much means we are getting close"

    • How do agents fit into OpenAIs long-term plans? o1 models, and all of it's predecessors will be the thing that makes agents actually happen.

    • "People get used to any new tech quickly, but this (agents) will be a big deal… People will ask an agent to do something that would have taken them a month, and it'll take an hour. Then they'll have 10x agents, then they'll have 1000x agents"

    • Sam, people think that technology makes a company, not true, it takes a lot of execution. "It doesn't excuse you from building a good business." "People are tempted to forget that with AI"

    • "Saying please/thank you to ChatGPT is a good thing to do, you never know"

    • As we move the world of agents, OpenAI will try virtual employees *

    • Why can't it sing, "we can't have it sing copyrighted songs. We want it to sing too, but it's nuanced to get it right. We really want the models to sing too."

    • "When will we get to 10T tokens?" - Sam, "Infinite context length will happen within the decade"

• Also from Open AI: Introducing ChatGPT Canvas

  • For the first time, we are fundamentally changing how humans can collaborate with ChatGPT since it launched two years ago. We’re introducing canvas, a new interface for working with ChatGPT on writing and coding projects that go beyond simple chat. Product and model features:

    • 1/ Ask for in-line feedback. With canvas, ChatGPT can better understand the context of what you’re trying to accomplish. You can highlight specific sections to indicate exactly what you want ChatGPT to focus on. Like a copy editor or code reviewer, it can give in-line feedback and suggestions with the entire project in mind.

    • 2/ Directly edit the model's output and select a specific area for targeted editing. You control your creative work on canvas. You can directly edit text or code.

    • 3/ Menu of shortcuts. There’s a menu of shortcuts for you to ask ChatGPT to adjust writing length, debug your code, and quickly perform other useful actions. You can also restore previous versions of your work by using the back button in canvas.

    • 4/ Use search with canvas for research writing! As we are moving towards the new paradigm of reasoning we are fundamentally evolving the chat interface into a more collaborative human-AI interaction. Today you can say “browse / use browsing to find XYZ on the internet and write a report in canvas”

• The foundations of future AI - The future of AI will be an ecosystem of models

  • Interesting article by Azeem Azhar and Nathan Warren on why they believe AI’s next chapter will be written by a diverse ecosystem of domain-specific foundation models, each mastering its own data domain

  • They note that while GPT-4, Claude and Gemini benefit from the abundance of data on the internet there are many areas that lack abundant data e.g. there are 10,000 rare diseases, but each affects fewer than 200,000 people.

    They argue that specialised domains require AI models that are:

    • (1) Broad enough: The model has been exposed to a wide variety of data, allowing it to learn general patterns and principles

    • (2) Specific enough: The model’s architecture and training process are tuned to leverage the data effectively to make accurate and useful predictions in specific tasks. Their training data needs to be focused enough on what domain knowledge is “useful”, yet broad enough to be expressive in their domain. This is especially important when there is not enough data to impose structure

  • Linked to this theme I also read The Next Big Data Gold Rush: Why Vertical AI Needs Specialized Data Providers Now

    • Sebastiaan Vaessen, Co-Founder of Coalition Capital makes the same case but from an investor perspective:

      • Industry’s focus will soon shift away from broad, horizontal AI applications toward more specialized, vertical uses cases tailored for specific industries or business needs. As AI continues to narrow its focus, the demand for high-quality datasets to support these targeted use cases will only increase. This shift presents a considerable opportunity for both investors and innovators.

• It’s the Year 2030. What Will Artificial Intelligence Look Like? - A collection of opinions on what AI will look like…

  • Here is one:

    • AI personal assistants to handle our lives (Amy Webb, chief executive, Future Today Institute):

      • By the year 2030, we will each have a Personal Large Action Model, or PLAM. These are advanced AI agents designed to replicate and emulate our unique decision-making processes. A PLAM will use data collected from the devices we wear (earbuds, continuous glucose monitors) and the devices we use (smart toilets embedded with sensors, digital wallets) to understand our likely behaviors and act on behalf of us.

      • Our PLAMs, which we will train through repeated use, will learn and adapt to our individual unique behaviors, actions, mood expressions and preferences, and they will invisibly and autonomously handle complex tasks, such as negotiating rates with CLAMs, or Corporate Large Action Models.

      • Here’s how that could look: As a business traveler, you have a coming trip from New York City to Munich, and there are no direct flights. Your PLAM would let airline carriers know it wants a deal and allow them to compete for your business. After brokering an itinerary optimized to what you specifically like (preferred seating configuration, just the right amount of time to make your connection, ability to accumulate points), your PLAM completes all of the tedious steps involved in actually purchasing the ticket: signing in to the airline and authenticating, entering your information, selecting which credit card to use, authenticating for payment, submitting payment details, entering the receipt details in your company’s accounting system and posting travel details to your calendar. With enough PLAMs deployed, yours could work to find you the perfect seatmate, too.

• Why Mark Zuckerberg thinks AR glasses will replace your phone

  • The Verge interview Mark Zuckerberg after the Orion showcase last week which I covered in my last edition

    • Mobile as a mid-point on the compute continuum: There’s this continuum of computing where, now, you have a mobile device that you can take with you all the time, and that’s amazing. But it’s small, and it kind of pulls you away from other interactions. Those things are not great.

      [We recognised that in ] the transition from computers to mobile, mobile was not…the end of the line.

    • AR glasses as the next step: The vision is a normal pair of glasses that can do two really fundamental things:

      • Holograms in the world to deliver this realistic sense of presence, like you were there with another person or in another place, or maybe you’re physically with a person, but just like we did, you can pull up a virtual Pong game or whatever. You can work on things together. You can sit at a coffee shop and pull up your whole workstation of different monitors. You can be on a flight or in the back seat of a car and pull up a full-screen movie theater. There’s great computing and a full sense of presence, like you’re there with people no matter where they are.

      • It is the ideal device for AI. Glasses are uniquely positioned for you to be able to let people see what you see and hear what you hear. They give you very subtle feedback where they can speak in your ear or have silent input that shows up on the glasses that other people can’t see and doesn’t take you away from the world around you. I think that is all going to be really profound. I had thought that the hologram part of this was going to be possible before AI. It’s an interesting twist of fate that the AI part is actually possible before the holograms are really able to be mass-produced at an affordable price.

    • How it might compare to the Meta Quest / Apple Vision Pro:

      • At each step along this continuum, from display list to small display to full holographic, you’re packing more technology in. Each step up is going to be a little more expensive and is going to have more constraints on the form factor. You’ll be able to do the simpler ones and much smaller form factors permanently. And then, of course, there are the mixed reality headsets, which kind of took a different direction, which is going toward the same vision…. [they have] all the compute we want, and.. [are] more of a headset or goggles form factor.” There are a bunch of uses where people want the full immersion. And if you’re sitting at your desk and working for a long period of time, you might want the increase in computing power you’re going to be able to get. But I think there’s no doubt that what you saw with Orion is the quintessential vision of what I thought and continue to think is going to be the next major multibillion-person computing platform. And then all these other things are going to get built out around it.

    • On AI agents:

      • Every business… just like they have a website, email address and a social media account today, I think that they’re all going to have an AI that helps them interact with their customers in the future, that does some combination of sales and customer support and all of that. I think all the creators are going to want some version of this that helps them interact with their community when they’re just limited by not having enough hours in the day to interact with all the messages that are coming in, and they want to make sure that they can show some love to people in their community.

• Newsom Vetoes SB 1047 - Zvi

  • California Governor Gavin Newsom vetoed SB 1047, a bill aimed at regulating large AI models and requiring developers to implement and disclose safety protocols

  • There have been lots of different takes on this and arguments on both sides. Zvi is very much the anti-veto side giving his view on why Newsom vetoed the bill. He argues that Newsom’s justification—that the bill is insufficiently comprehensive, focusing only on large models and potentially giving a false sense of security—is unconvincing. The veto was likely motivated by concerns about the bill's impact on industry and innovation or by political calculations rather than Safety concerns. Newsom’s declared preference for regulating AI through use-based restrictions rather than model-based ones is criticized as ineffective in addressing existential risks posed by advanced AI systems. Zvi warns that focusing on specific AI uses, rather than the development of potentially dangerous models themselves, will lead to burdensome and ineffective regulations, similar to those in the EU, that could stifle innovation without mitigating the most serious risks. Zvi by contrast calls for a model-focused regulatory framework, such as SB 1047, as a better path forward for balancing innovation with safety.

Interesting AI use cases of the week:

• Facial Recognition Tech on Meta's Smart Glasses

  • Two Harvard students have created I-XRAY, facial recognition software built on top of Meta’s Smart Glasses. It can identify faces and gather personal information like addresses and phone numbers. It uses LLMs to understand, process, and compile information from diverse sources–inferring relationships between online sources, such as linking a name from one article to another, and logically parsing a person’s identity and personal details through text. This synergy between LLMs and reverse face search allows for fully automatic and comprehensive data extraction that was previously not possible with traditional methods alone.

  • This is a topic that big tech has avoided due to safety concerns but it seems like people will be able to do it by themselves…

  • I highly recommend you follow the link and watch their demo video for yourself!

• Meta just announced Movie Gen

  • Meta just announced Movie Gen, and it’s already blown away Sora, Runway, and other video generation tools!

  • It’s probably the biggest release of the year in Gen AI.

    • Use simple text inputs to produce custom videos and sounds

    • Edit existing videos

    • Transform your personal image into a unique video

    • Generate audio effects

• DeepMind and BioNTech build AI lab assistants for scientific research

  • Google DeepMind and BioNTech are building AI lab assistants to help researchers plan scientific experiments and better predict their outcomes as companies race to find specialised applications for energy and data-intensive artificial intelligence models. It should also help make unexpected connections more easily.

  • German drugmaker BioNTech and its London-based AI subsidiary InstaDeep has: designed a specialised AI assistant known as Laila with a “detailed knowledge of biology” built on top of Meta’s open-source Llama 3.1 model…. the AI agent could automate routine scientific tasks in experimental biology, such as analysis and segmentation of DNA sequences, and the visualisation of experimental results. Laila could connect to lab devices and monitor ongoing experiments or tasks being performed by robots, with the assistant detecting a mechanical failure from a BioNTech machine during a live demonstration.

    • “We see AI agents like Laila as a productivity accelerator that’s going to allow the scientists, the technicians, to spend their limited time on what really matters,” - Karim Beguir, CEO of InstaDeep. While recognising that rivals such as DeepMind could also build AI assistants, Beguir said having InstaDeep’s technology “under the same roof” as BioNTech’s expertise in biology was an “accelerant” for implementing AI and “unique” in the pharmaceutical sector.

  • The wave of AI innovation in science has so far focused on predicting new and useful drug candidates. However, the bottleneck in bringing new treatments to market remains performing experiments in the real world, which is the gold standard in scientific research. The goal of an AI research assistant would be to simplify this process by more effectively planning experiments, for example by selecting the most promising from among a set of possible experiments.

• The business of football: the big data arms race | FT Scoreboard

  • A technological revolution is underway in football, as team owners turn to the latest data analytics and AI to gain a competitive edge in the battle for talent. As more professional investors buy into the sport and rules on spending get tighter, future champions will increasingly be created away from the pitch by teams of software engineers.

  • The video profiles AI.IO, a sports performance AI company working with football clubs like Chelsea and Burnley

• Alibaba, Nvidia collaborate on advanced autonomous-driving solution, computing services

  • This collaboration marks the first time Alibaba Cloud’s AI models get integrated into Nvidia’s Drive automotive platform

  • This marks the first integration of Alibaba Cloud’s large AI models into Nvidia’s Drive automotive platform, which major Chinese electric vehicle (EV) makers – including Li Auto, Great Wall Motors, Geely Auto’s Zeekr Intelligent Technology and the EV unit of Xiaomi – are using to power their next-generation fleets.

• How AlphaChip transformed computer chip design

  • Google Deepmind explains how their AI method has accelerated and optimized chip design, and how its chip layouts are used in hardware around the world

• Japan's Shimano gives cyclists an easier ride with AI gear shifter

  • Japanese bicycle parts maker Shimano plans to launch an AI gear shifting system. The device results from two decades of development following the release of a bike in the early 2000s with electronically controlled shifting. It will customize the tempo and speeds based on the rider's habits and have memorised the bike rider’s patterns for test ride data including the speeds and the inclines taken by the cyclist in order to choose the right gears at the right time. Target market: mass consumer, same as e-bikes. Expected to hit the market as soon as 2025

Interesting Start-Up news of the week:

• Over the last 4 years, we’ve been building an AI model that accurately predicts seed-stage startup outcomes. We ranked all 50 factors from most to least predictive. Top 5 below.

  • Seed runway. The size of a seed raise had a monumental impact on a startup’s chances of surviving long enough to find PMF.

  • Founder personality features (we use NLP to assess founder traits such as neuroticism, agreeableness and extraversion) were the second most predictive factor.

  • Time between incorporation and first raise was also influential; companies that waited more than 3yrs between founding and first raise were substantially less likely to raise future Series-A or Series-B rounds.

  • Number of founders was also highly predictive, with more founders increasing the chances of future raises (although benefit is marginal beyond 3 founders).

  • Prior exp at a startup was a key predictor, although this didn’t have to be in a founder role.

• Vinod Khosla’s Founder Mode

  • 3 quick lessons from Vinod in the traits he believes are common to all the best founders:

    • Flexibility and stubbornness. The best entrepreneurs are “flexible in their tactics but obstinate in their vision,” according to Vinod. They must be willing to adapt to the game on the field but constantly keep their end goal in mind. Sam Altman and Patrick Collison are founders Vinod believes exemplify these traits. 

    • Discerning learners. To survive as a startup CEO, you must be or become a learning machine. The job is so varied that it requires you to level up, day after day, week after week. Exceptional founders are not only fast learners; they are discerning ones. They have no problem throwing out bad ideas, even if presented by a perceived authority figure. They’re equally happy to accept good ones, no matter their provenance. 

    • Some problems are too important. There are dozens of frameworks and heuristics venture capitalists use to analyze potential investments. For some opportunities, Vinod asks himself a simple, but powerful question: how crucial a problem is this startup solving? Over the years, the Khosla Ventures boss has backed companies principally because they were “too important not to do.” Think nuclear fission, AGI, or plant-based meat

• LinkedIn Top Startups 2024: The eight European companies on the rise

  • I’m less interested here by the Start-Ups themselves but rather Linkedin’s methodology: They leverage exclusive LinkedIn data across four key pillars: employee growth; jobseeker interest; engagement within the company and its employees; and how well these startups have pulled talent from Top Companies.

    • I’m waiting for the launch of Linkedin Ventures! Although perhaps Reid Hoffman already leverages this for his own angel investments!

A Start-Up that I like:

• Longshot - reinventing space launch

  • The team are building the world’s largest hypersonic accelerator to supply the orbital economy and provide responsive hypersonics testing

  • To achieve a future where humanity has access to outer space we need to put gigatons of raw materials and equipment into orbit. Humanity currently launches cargo with the same vehicle as humans - rockets. Cargo can withstand much higher accelerations. This is why we are focused on kinetic launch. The lowest possible cost for putting mass to orbit with simple and reliable technology - guns.

  • Longshot investors include Tim Draper and Sam Altman, and has been awarded $3M across four contracts from the US Air force focused on hypersonic testing.

Science news of the week:

• Why do obesity drugs seem to treat so many other ailments?

  • From alcoholism to Parkinson’s, scientists are studying the mechanisms behind the broad clinical potential of weight-loss drugs.

  • The article discusses how GLP-1 receptor agonists, initially developed as weight-loss drugs, show potential in treating a wide range of conditions, from addiction and cardiovascular disease to neurodegenerative disorders like Parkinson's. These drugs, like semaglutide, influence neural pathways related to reward and inflammation. Their broader effects beyond weight loss are being explored in hundreds of clinical trials, though scientists still don't fully understand the mechanisms behind some of the observed benefits. However, concerns remain about long-term use and side effects.

Sports business news of the week:

• Pentagon contracts for $96M in Oura smart rings, services

  • Human performance technologies originally designed for sport are proving to have much further reaching use cases!

  • “This contract is to provide the Defense Health Agency (DHA) Wellbeing Office delivery of Ouraring Inc., biometric sensor devices; data analysis including monitoring of physiological stress, recovery, resilience, and wellbeing indicators; individualized biometric data visualization; aggregate wellbeing visualization for the agency; and content delivery of wellness-related insights and training”

• Lalit Modi - Founder & Architect of the IPL - publishes his analysis of The Hundred

  • Modi argues that the sell-side financial model projections of The Hundred’s central league and team performance are wildly optimistic and misrepresent the actual fundamentals of the Hundred: particularly in the areas of Media Rights, League & Team Sponsorships and Profitability

  • Modi provides more details on his analysis in a conversation on the Unofficial Partner podcast: “In a market that is flat, going from 5.7m (pounds) to 32m is again wishful thinking. Have they demonstrated that the sponsorship revenue is going up year after year? It isn’t. The guaranteed revenue is 5.7m, going to 6.3m next year. How are the jumps going to take place to get it to 32m? Then, you come to the major issue. They have 2.1m pounds today for international rights up to 2032, it constitutes about 47 per cent of the total revenue The Hundred is going to make. For the IPL, international rights are about 3 per cent. The NFL has a similar number, they have negligible international revenue. They say India is the Mecca for cricket. Yes, but for Indian cricket. People pay and watch Indian players. Then, they go on to say that the American rights are going to go up phenomenally. That’s not been proven for anybody, including the IPL. They’re saying the World Cup that took place is a demonstration of that. They’re saying the Olympics coming in will further demonstrate that. Let’s go back to the [T20] World Cup that just finished. Who watched that? Only India matches were watched. And that too India-Pakistan and some others. People are pegging their coats on the World Cup having done well. It hasn’t. India was participating and India made it to the final. You saw the debacle in the West Indies in 2007, when India were out. It had a major impact on the viewership and revenues. You cannot base your international rights on that.”

• An interesting post by George Pyne (Founder & CEO of Bruin Capital)

  • George discusses why the decision of the MLS to partner with OneFootbal is a clever play and the latest example of a sports league capitalising on the intersection of media, technology and data through equity investment

  • He notes that the relationship allows MLS to draw deeper and more valuable insights into how fans engage with its IP – particularly the league’s match highlights, which are traditionally hard to discern quality audience insights from.

  • This is important in a world in which Sports fans are increasingly relying on game highlights as their primary means of consumption. Per Morning Consult, 58% of Gen Z fans prefer watching highlights over full games, compared to 46% of millennials. This signals a generational shift towards shorter, on-demand formats via platforms like YouTube, Instagram, Twitter, and TikTok. Data from Statista also suggests that 49% of all sports fans, regardless of generation, prefer consuming sports content through highlight compilation videos, particularly in Asia, where time zone differences can make it challenging to watch live games.

    • This is a key reason why it’s hard to believe that the European Super League is gone for good (you can guarantee the best players in the world from across Europe will be playing on the same pitch every year - great for highlights) and don’t forget the continued rise of Gerard Pique’s Kings League which raised another 60m EUR back in May…

World news of the week:

• Preliminary Milei Report Card by Scott Alexander

  • Scott provides an initial take on Javier Milei’s progress so far as President of Argentina. Here is his conclusion:

    • When Javier Milei took office, he promised to do shock therapy that would short-term plunge Argentina into a recession, but long-term end its economic woes. He has fulfilled his campaign promise to plunge Argentina into a recession. Whether this will long-term end its economic woes remains to be seen. I think he gets credit for some purely political victories (completing the budget cuts he said he would complete), for decreasing inflation, and for improving the housing market. But in the end, history will judge him for whether his shock therapy eventually bears fruit. I don’t think that judgment can be made yet, and I don’t see many economists eager to go out on a limb and say that there are strong signs that his particular brand of shock therapy will definitely work/fail.

  • For further reading on Argentina this week, Samo Burja in Why Argentina’s Social Democracy Failed. Reiterating the importance of a political live player like Milei, Samo argues that Argentina’s success depends on whether Milei can undertake the deep political, economic, and institutional reform neededed to overcome Argentina’s fundamental problem: its’ need to industrialize more heavily and specialize in particular globally-competitive sectors, upsetting the foundation of the country’s political order since 1946.

• The Creation of Lebanon After The First World War

  • Given the sad news in Lebanon right now I realised it was another country I knew far too little about, a country in which my grandfather lived in the 1950s when Beruit was known as “the Paris of the Middle East”

  • This YouTube video provides a fascinating primer on the events that led to and shaped the creation of Lebanon after the First World War.

• Three Days in North Korea

  • A fascinating description of a three-day guided trip to North Korea. Here is one excerpt:

    • “From the bus, we could see hundreds of people kneeling in a large square, praying and bowing. It was one of the most bizarre scenes I have ever witnessed. Apparently, they were rehearsing for the party's anniversary celebration—at 4:30 in the morning”

China news of the week:

• China sells more cars powered by electricity than traditional fuel

• The $1.7 Billion Takeover Brawl Fueled by a Fear of China

  • Concerns about Beijing’s mineral-resources dominance hang over a deal to control the world’s largest zinc smelter

  • MBK, an Asian private equity firm with >$30bn AUM, is seeking to purchase a controlling stake of up to 47.7% of Korea Zinc. However, MBK owns a handful of Chinese companies and has a small contingent of Chinese investors but despite pledging not to sell Korea Zinc to the Chinese, Korea Zinc argues the risk is still there and it is fighting back against what it classes as a hostile takeover attempt of a resource of national security influence.

  • The showdown over Korea Zinc illustrates how the mere potential of technology shifting to Beijing can complicate dealmaking in once-sleepy corners of the global supply chain. Despite Western efforts, China’s advantage in minerals has only expanded, in everything from nickel to cobalt to lithium

    • Update: Korea Zinc Teams Up With Bain Capital to Thwart Takeover Bid - Korea Zinc said it could spend more on the share buyback, if needed, in order to stop the MBK-led takeover attempt

India news of the week:

• Swiggy IPO

  • Swiggy IPO: Food delivery giant Swiggy set for IPO - targeting a valuation of $15 billion after it was valued at $10.7 billion in its last funding round in 2022 led by asset manager Invesco.

  • The quick commerce sector in India is witnessing significant growth. India’s food delivery market is a duopoly with Swiggy and Zomato together commanding over 90% of the industry, while the market is estimated to grow to ₹2 lakh crore by 2030. Here are four key stats:

    • The number of customer transactions over three years has expanded from 243,000 to 323,000 to 393,000, the number of users from 35m to 43m to 47m.

    • The number of gig workers making deliveries has almost doubled since 2022 to 457,000 while their payments have remained relatively flat at 58 rupees, or $0.69, for each order. They deliver on average 463 orders each month.

    • The number of restaurants whose orders were delivered through Swiggy has grown from 129,000 in 2022 to 224,000 at the end of June.

    • Four warehouses in 2022 have expanded to 50. It has begun building “dark stores”, meaning ones that have no retail customers but are packed with common items and positioned to enable fast delivery. There were 12 in 2022; there are 538 at the end of June.

• Campa Cola & Reliance

  • A popular brand in India in the 1970/80s after Coca-Cola was forced to leave because it refused to give up its formula, it was acquired by Reliance, India’s biggest private company in 2022.

  • Now it is being relaunched in the Reliance way, which means cheap prices and wide distribution, to take on Coke and Pepsi…. The American giants are no strangers to low-end competition and Reliance, notwithstanding its dominance of oil refining, telecoms and the gargantuan weddings of its promoter’s children, seems, anecdotally at least, to have never developed the same level of trust among Indians as Tata, Bajaj, TVS, HDFC Bank, Hindustan Unilever and a handful of other major companies operating in India. That is because its success has relied on other factors. The refining business depends on engineering and scale and ensuring cheap supplies of oil from controversial sources are available and not blocked by countries with opposing foreign policies. The telecom business succeeded with the broad public based on very cheap prices. In theory, fizzy drinks are just soda, sugar and a bit of flavouring. Reliance will certainly provide that at a very low price and it has hired at least one former Coke bigshot. That would seem to be a compelling package. But others have failed pursuing similar strategies. To succeed, Reliance will have to become genuinely liked by the public. That will be an interesting test. 

  • As India continues to rise I expect we will see much more attempts by domestic Indian companies to gain market share vs. international incumbants

UK news of the week:

• Eight Scientists, a Billion Dollars, and the Moonshot Agency Trying to Make Britain Great Again

  • Against the backdrop of economic and political doom and gloom in the UK, a genuine positive was the establishment of ARIA - the Advanced Research & Innovation Agency. Kept at a careful (legally secured) distance from government medling it is nonetheless a government-funded “start-up” with a healthy budget designed to invest in high-risk high-reward scientific endeavours that the free market might ordinarily choose to pass on given the risk profile. Modelled on ARPA (later DARPA) this article is a great overview of the programme, and how and why it was set up.

  • As a side note, I want to congratulate Dominic Cummings for helping push this through - I remember reading at School his 2013 blog post where he called for a high-risk/high-reward agency focused on energy projects that “operates outside all normal bureaucratic systems,”. This was a topic he returned to many times in the following years and was a deal breaker in his decision to become Boris Johnson’s chief advisor… “double the science budget, avoid a second referendum, overhaul Whitehall, and create a breakthrough funding agency modelled on ARPA”. While he never quite managed the Whitehall overhaul (I say watch this space), he did get ARIA over the line.

    • I remember being hugely excited back in 2019 when I spotted ARIA’s soft launch on page 40 of the Conservative Party Manifesto! The bill then setting up ARIA became law in February 2022!

• Foundations: Why Britain has stagnated by Sam Bowman, Samuel Hughes & Ben Southwood

  • Hugely important essay on both why Britain has stagnated and what it needs to do to unlock prosperity (spoiler: it’s not focusing on taxing private school fees and further clamping down on non-doms)

  • Byrne Hobart sums up the essay excellently:

    • “Economic growth is subject to technological and material constraints, but lack of growth over long periods is a choice. The theme of this piece is that the UK has an opportunity to achieve incredible outcomes by doing the basic things right—building houses (France has 1.2x as many per capita), getting cheaper energy (the US consumes almost 3x as much per capita), getting to parity with other countries on infrastructure (Spain gets twenty times as much railroad mileage per dollar). Technology-driven productivity growth matters a lot in the long run, but sometimes the limiting factor behind it is that people can't take promising-but-low-paid jobs at early-stage tech startups because nobody's allowed to build the housing they need to work there.”

Academic paper(s) of the week

• Like owner, like dog – A systematic review about similarities in dog-human dyads

  • Yes, owners do look like their dogs!

    • This systematic review examines empirical evidence supporting the anecdotal assumption, that dogs look like and behave like their owners. To this end, we investigated 15 studies with the aim of testing that: (1) Owners and their dogs resemble each other in appearance and (2) owners and their dogs have similar personalities. Aggregation of the results supports evidence for both hypotheses. In terms of external similarities (i.e., appearance), participants were able to correctly match photos of dog-owner pairs, with one study even demonstrating that the eye area of owners and dogs was sufficient for accurate pairings. Further significant attributes included the correlation of weight of dog and owner and the hair/ear length similarity between women and their dogs. The included studies explain this by the owners' choice of a similar looking dog, possibly influenced by psychological mechanisms such as the Mere Exposure Effect or a Preference for Self-Similarity. Regarding internal similarities (i.e., personality), correlation of various personality dimensions was found. Most studies support the existence of similarities on the dimensions of extraversion and neuroticism. Discussion of potential reasons for these findings proposed a development of similarity over time, due to effects such as Emotional Convergence, Co-Regulation, and Reinforcement.

• If the face fits: predicting future promotions from police cadets’ facial traits

  • Facial traits are the primary driver of subject perceptions of leadership ability, and those perceptions successfully predict promotional success later in the cadets’ careers. When selecting for leadership potential based on police cadet photographs, respondents predict correct promotional choices at levels well above chance as measured by an AUC score of .70. Further, respondents’ evaluations successfully discriminate both between no promotion and lieutenant promotion, and sergeant versus lieutenant promotions.

  • More evidence for the beauty premium

• Can Pollution Markets Work in Developing Countries? Experimental Evidence from India*

  • Short answer yes they can…. and as Alex Tabbarok notes on top of the study: “there are significant productivity gains to reducing air pollution which would make these benefit to cost ratios even higher. Less pollution can mean more health and more wealth”

Culture news of the week:

• No, Culture is Not Stuck - You just can't see what it's become

  • While the idea that culture is stagnating is becoming increasingly common (consider Film & TV (rise of repeats, focus on specific popular IP e.g. Marvel/DC superhero , Fashion and Music e.g. old songs killing new songs, tiktokification of music), Katherine Dee argues that perhaps the problem is the way we define “culture”. She argues that new forms of cultural expression, particularly social media personalities and TikTok content, are often overlooked or dismissed by critics because they don't fit traditional notions of art or performance. Content creators represent continuous, evolving personas across platforms, which can be seen as a type of art where audience engagement plays a crucial role. This form, while innovative, is often undervalued because of the sheer volume of low-quality content. Similarly, TikTok, often criticized as trivial, showcases creative innovation, particularly in comedy, likened to digital vaudeville. The dismissal of these mediums overlooks potential works of genius within them.

• The Other Place: A new play by Alexander Zeldin, After Antigone

  • On Monday evening I watched The Other Place starring Emma D’Arcy (House of the Dragon), Alison Oliver (Saltburn), Tobias Menzies (The Crown) and Nina Sosanya (Love Actually)

  • The Other Place’ is a rewrite of Sophocles’s ‘Antigone’.

    • Two sisters reunite on the anniversary of the death of their father. Their uncle has remodelled their family home, in an attempt at a fresh start. But one sister’s sudden reappearance threatens to shatter this fragile idyll as she demands justice for the pain she carries. Amid the debris and the new extension, guilt, grief and greed battle it out in the family’s competing dreams of their future. When we are faced with the suffering of others, even those closest to us, can we look away?

  • Director Alexander Zeldin has said that :

    • “Antigone is a play about the aftermath… It’s not about someone resisting arrest or something: it’s [about] two forms of grief… it’s a modern tragedy around one person wanting to erase the past and one person wanting to preserve it…We have radically different ways of dealing with grief…There are situations in life when there is no resolution possible and acknowledging that has a great value.”

  • From Emma D’Arcy:

    • The Other Place is about a blended family that reunites 10 years after the death of a patriarch, and what ensues is a kind of territory war fought over the past. What we’re exploring is the gap between a story and the truth, and the rot that’s able to grow there. Sophocles was made a general after he wrote Antigone, because it was considered that he had a complex understanding of the workings of power – I find that really extraordinary.

  • I have so many thoughts on this play and may well update this section of my blog as time passes, but my overall view was that it was brilliant (5 stars). Obviously the play’s themes are very interesting but also: The music and staging were fantastic; The casting was superb, I found the play particularly disturbing (in a good way if you can say that!) because of just how authentic, relatable and familiar many of the characters felt (before things take a much darker turn!); and the pacing of the play was spot-on, building to this horrific crucendo at the end.

  • Side-note 1: As within The Other Place, I am so impressed by the ability of directors to include frequent comedic episodes within art of a very serious content matter without detracting at all from this seriousness. It’s take a very special touch. Bong-Joon Ho, the Korean film director, is also brilliant at this.

  • Side-note 2: I couldn’t help but reflect afterwards on Ted Gioia’s How to Read Greek Tragedy in a Netflix World. Here are some excerpts:

    • Greek tragedy teaches us that there’s no final rescue or explanation in the last scene—and certainly nothing resembling human justice. When you want to assess a person’s life, you don’t judge it by the cause of death in the obituary. Life doesn’t work that way. Only stories do—and the ancients were rightly distrustful of those types of stories.

    • More than 90% of Hollywood stories have happy endings. (I’ve heard some estimates that it’s as high as 99%—but we lack reliable measures.) And the public prefers this—even when the resolutions are improbable and unrealistic. This is why Netflix has never released Oedipus: The Miniseries—and never will. Greek tragedy is too disturbing to the modern sensibility. These stories remind us of the shortness of life, the unpredictability of human affairs, the risks we take when we embark on any ambitious personal venture, and other painful truths.

    • By contrast, even The Book of Job—the gloomiest story in all holy scripture—serves up an implausible happy ending in the final scene. Only the Greeks were courageous enough to dispense with such comforts and face destiny straight in the face, never blinking. None of these stories would get funded by Hollywood today, not even for a test pilot episode. And that’s exactly why it’s so mind-expanding to encounter these dramatic scenarios.

    • That’s why René Girard developed his theory of sacrificial violence from studying these ancient tragedies. He saw that the death of the victim is both an end and a new beginning. Trying to balance out the costs and benefits may be beyond human metrics, but that tension gives depth and power to the story.

• Conversations with Tyler: Kyla Scanlon on Communicating Economic Ideas through Social Media (Ep. 222)

  • This podcast is a broad-ranging discussion between Kyla Scanlon, who aims to bring economics education to a larger audience through social media, and Tyler Cowen. I just wanted to call out one part that I found particularly interesting:

    • COWEN: Is celebrity dead?

      SCANLON: Oh, [laughs] yes.

      COWEN: And do you need celebrity? You said it’s all I need, but in fact, you have however many TikTok and YouTube videos.

      SCANLON: Yes, I know. There’s always a little bit of hypocrisy sometimes, I suppose. I think celebrity is dead. I put that down because I was curious what are your thoughts on that as well. I think people look toward influencers as celebrities now. The celebrities themselves are just sort of these figments of movies. An influencer is somebody who’s going to be your “best friend,” somebody that you tune into every aspect of their lives, and they give you access to every aspect of their lives. I think that far supersedes everything that we’ve ever gotten from a celebrity.

• Apple spent billions making movies for theatrical release, it didn’t really work

  • Apple will send more movies straight to streaming as it switches up its plan to pump $1 billion every year into theatrical releases, having spent more than $20 billion on a string of hit-and-miss movies and TV shows over the last 5 years.

    

Miscellaneous

• Why So Few Matt Levines?

  • Why are popularizing educational newsletter-frequency writers of important fields like Matt Levine for finance so rare? Because most fields are too slow or ambiguous, and writers of the right combination of expertise, obsession, & persistence are also rare.

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please crosscheck with my actual substack: https://substack.com/@thespecialistgeneralist. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

Weekly Recap #2 explores the rapid evolution of AI and its impact on various aspects of life, from education and healthcare to storytelling and the job market. The hosts discuss the changing landscape of the tech world, with the rise of vertical apps catering to specific needs and the potential disruption of traditional platforms. They highlight scientific breakthroughs like creating viable eggs from male mouse cells, which could revolutionise reproductive health, and the study of Greenland sharks for insights into longevity. The conversation also touches upon the economic rise of India as a tech hub, the prevalence of gloom in pop culture as a reflection of societal anxieties and the importance of storytelling in a world saturated with information. They emphasise that while these advancements and shifts can be overwhelming, human ingenuity and adaptability, as seen in various fields, offer hope for navigating the future.

Important: This is the podcast version of my weekly substack created by Google NotebookLM. As an AI-generated podcast, it may contain inaccuracies. If in doubt then please feel free to crosscheck with my actual substack: https://substack.com/home/post/p-149567467?r=4xlcq&utm_campaign=post&utm_medium=web. As a final disclaimer, I would say that the conversational style of the podcast is not exactly representative of me! With that out the way, I hope you enjoy the episode!

The "Weekly Recap #1" by Oliver Wilson spotlights key trends across technology, economics and AI. It examines the possibility of uncovering lost ancient texts through technologies like digital unwrapping. The recap also explores the contrasting urban development models of "producer cities" and "consumer cities" in developing countries and discusses the European Union's struggle to maintain global economic competitiveness against the U.S. and China. Additionally, it provides insights into the success strategies of YouTuber MrBeast, emphasizing data-driven optimisation and prioritising results. Lastly, the recap highlights notable AI advancements, including OpenAI's GPT-o1 and Google Research's bioacoustic model HeAR.

1. Top things I’ve read this week

2. AI news of the week

3. Interesting AI use cases of the week

4. A Start-Up that I like

5. Science news of the week

6. Sports business news of the week

7. China news of the week

8. India news of the week

9. UK news of the week

10. Academic paper(s) of the week

11. Culture of the week

Top things I've read this week:

• The Intelligence Age by Sam Altman

  • Sam Altman on the progress AI will generate (Deep Learning worked). It's always important to listen when one of the most pioneering figures of our age speaks so here he is in his voice:

    • "It won’t happen all at once, but we’ll soon be able to work with AI that helps us accomplish much more than we ever could without AI; eventually we can each have a personal AI team, full of virtual experts in different areas, working together to create almost anything we can imagine. Our children will have virtual tutors who can provide personalized instruction in any subject, in any language, and at whatever pace they need. We can imagine similar ideas for better healthcare, the ability to create any kind of software someone can imagine, and much more. With these new abilities, we can have shared prosperity to a degree that seems unimaginable today; in the future, everyone’s lives can be better than anyone’s life is now. Prosperity alone doesn’t necessarily make people happy – there are plenty of miserable rich people – but it would meaningfully improve the lives of people around the world."

    • As expected given what we have heard from former OpenAI employees recently he is light on the risks bar saying that there will be "very complex and extremely high-stakes challenges". But the crucial insight into how Sam thinks is that he believes "the upside is so tremendous that we owe it to ourselves, and the future, to figure out how to navigate the risks in front of us".

    • As Zvi comments: note the contrast:

      •  Sam Altman (2015): "Development of superhuman machine intelligence is probably the greatest threat to the continued existence of humanity."

      • Sam Altman (2024): "this technology can cause a significant change in labor markets (good and bad) in the coming years, but most jobs will change more slowly than most people think."

    • Let's hope we can navigate these risks for all of our sakes.

• The Comfortable Queen and the Conflicted Country - Vance, Kamala, and the malleability of identity by Richard Hanania

  • Richard Hanania explores the malleability of identity, using Kamala Harris and J.D. Vance as case studies. He delves into how ambition can shape not only one’s positions but also the way individuals recast their life narratives to adapt to evolving political climates.

  • For Hanania, Identity politics, as manifested by figures like Harris and Vance, is less about fixed categories like race or class and more about how elites choose to instrumentalize those categories to navigate the political landscape. The more adaptive, like Harris, manage this with ease, while the conflicted, like Vance, struggle in environments that demand incoherence. The malleability of identity, in Hanania’s view, reveals that political discourse in the U.S. is increasingly defined by the agility with which leaders pivot to fit the moment, rather than a coherent or fixed ideology.

• On Armenia by Maxi Gorynski

  • Maxi goes full Matt Lakeman and provides a really interesting History of Armenia along with his own experience visiting

  • Armenia is a country that I know should know more about. When I visited Israel & Palestine two years ago I was very interested to find an Armenian Quarter of Jerusalem’s Old City and to learn that the Armenian Apostolic is one of three Christian denominations (along with Catholic and Greek Orthodox) to own and administer the Church of Nativity in Bethlehem. After learning that Armenia was the first nation ever to declare Christianity as its national faith (301 AD) this is no longer so surprising! In any case, this article definitely made me keen to visit and I'll make sure to re-read before such visit.

• Book Review: On the Edge: The Fundamentals by Zvi

  • Great series of reviews of Nate Silver's latest book On the Edge

  • On the Edge is a book about those living On The Edge, the collection of people who take risk and think probabilistically and about expected value. It centrally covers poker, sports betting, casinos, Silicon Valley, venture capital, Sam Bankman-Fried, effective altruism, AI and existential risk. Collectively, Nate Silver calls this cultural orientation The River.

  • It is contrasted with The Village, which comprises roughly the mainstream mostly left-of-center institutions, individuals and groups that claim that they are The Experts and the Very Serious People.

    • Riverian mode in a nutshell:

      • ⁠⁠In this mode, you think in terms of probabilities and expected value (EV). You seek the most accurate possible model of the world, including what actions are how likely to lead to what results. You the riverian look to make the best decisions possible. You are not afraid of risk, but seek to take only the good risks that are +EV.

      • You can then be somewhat risk-averse when making decisions, risk-neutral or even risk-loving. All riverians have weaknesses, ways they systematically mess up. The key is, you accept that risk is part of life, and you look to make the most of it, including understanding that sometimes the greatest risk is not taking one.

      • Riverians are the Advantage Players of life.

      • They want life to be about everyone making good decisions. A True Riverian learns to inherently love a correct play and hate a mistake, in all contexts, from all sides that are not their active opponents. They want those good decisions and valuable actions to be rewarded, the bad decisions and destructive actions punished. They want that to be what matters, not who you know or who you are or how you play some political game.

      • Riverians hate being told what to do if they don’t think it will help them win. They despise when others boss them around and tell them to do dumb things, or are told to copy what others around them do without justification.

      • The River is where people focus on being right, taking chances and doing what works, and not letting anyone tell them different.

• Daniel Yergin – Oil Explains the Entire 20th Century - The Dwarkesh Patel podcast

  • Super interesting conversation on how “Oil Explains the Entire 20th Century, the rise of America, WW1, WW2, secular stagnation, the Middle East, Ukraine, how Xi and Putin think, and basically anything else that's happened since 1860”

  • One important takeaway is the importance of energy security through diversification: Yergin reinforces the idea that energy security comes from diversification, echoing Churchill's view that "safety lies in variety." This philosophy is evident in how nations like China and India pursue a mix of fossil fuels, renewables, and nuclear energy to mitigate risks from over-reliance on any single source of energy. China’s strategic shift to electric vehicles, for instance, is not just an environmental decision but a hedge against its vulnerability as an oil importer. By diversifying into electric mobility and renewables, countries are trying to secure more resilient energy futures amid volatile global markets.

AI news of the week:

• OpenAI pushes ahead with for-profit plans and talks to give Sam Altman a stake - The FT

  • Sam Altman’s battle with OpenAI’s board is now complete (remember the original board did not trust him and wanted him out).

    • Dylan Matthews: Remember when OpenAI's nonprofit board was like "this Altman guy is constantly lying to us and doesn't seem like he takes the nonprofit mission at all seriously" and people called them "clods" and mocked them? It's fun that they were completely right.

    • Benjamin De Kraker: Remember: Altman previously testified to the U.S. Senate that be wasn't doing it for the money and didn't have equity.

    • Eliezer Yudkowsky: Can we please get the IRS coming in to take back control of this corporation, avert this theft of 501c3 resources, and appoint a new impartial board to steward them?

    • Igor Kurganov: If you fire everyone who joined your non-profit, does it auto convert to a for profit?

  • Also this week, Mira Murati graciously leaves OpenAI. Interesting timing with the move to for-profit status! Also interesting was her role in the events of last November, when she temporarily was willing to become CEO

  • After a raft of Safety-related departures, the not-for-profit structure (that clashed with Sam) is also leaving.

  • Presumably, this will serve as a warning to others. You come at the king, best not miss. The king is not a forgiving king. Either remain fully loyal at all times, or if you have to do what you have to do then be sure to twist the knife.

    

  • Also: Apple No Longer in Talks to Join OpenAI Investment Round

• Meta showcased its new AI infused R glasses Orion

  • The Orion glasses are a prototype (sadly with no direct route to commercial sale) of what Meta thinks will replace your smartphone in the coming years. It's pretty early days still, which is a lot to say for the billions of dollars Meta has poured into them and the metaverse, but it's a seriously cool product and I'm glad Meta's showed us what it has been working on… I look forward to what comes next!

  • You can watch a cool first review from the Verge here:

  • https://x.com/RihardJarc/status/1839014234266755473

    • For a more technical overview directly from Mark Zuckerberg featuring Jensen Huang saying how cool the product is!

  • Also related Autism Capital on X discussing what they think the killer AR apps will be:

    • Maps overlay for directions

    • HR and vital signs for exercise

    • A beacon icon like in Halo to find friends at a festival

    • A simple name tag app so you never forget people you meet.

    • Dating app where if you wink at someone and if they wink back, you both light up green and you’re good to go. No BS. Basically IRL swiping.

    • Being able to leave virtual post it notes on real life objects so when you revisit that space and item you can have access to quick information

• Enterprise Philosophy and The First Wave of AI by Ben Thompson

  • Ben Thompson at Stratechery is always the place to go for interesting technology strategy insights. He provides an interesting overview of the the first Wave of AI and where it might go next (all in the context of Enterprise philosophy). I recommend reading it in full but here is a summary:

    1. Tech’s Two Waves: The first major tech wave transformed back-office operations, automating jobs like bookkeeping and accounting. The second wave, personal computing, empowered individuals to use technology for personal and business tasks.

    2. Two Tech Philosophies:

       • Google and Facebook: These companies focus on automating tasks, doing work for users (e.g., AI-driven suggestions, driving).

       • Apple and Microsoft: Their approach is about enhancing human productivity, giving users tools to do work more efficiently.

    3. Microsoft’s AI Vision: Microsoft’s “Copilot” integrates AI into workflows to assist workers rather than replace them. However, the success of Copilot depends on changing how employees work, which may be a difficult cultural shift for organizations.

    4. Agents vs. Copilots: Salesforce is betting on autonomous agents—AI that replaces human work entirely. This contrasts with Copilot’s approach, which aims to augment human productivity rather than replace workers.

    5. Enterprise AI: The big opportunity for AI lies in large enterprise implementations, where the focus is on replacing entire processes, not just making employees more productive. This will require years of integration and data preparation, much like the early days of mainframe computing.

    6. Palantir’s Role: Palantir is poised to play a key role in enterprise AI due to its focus on data integration. Their approach is essential for making AI work at scale in complex corporate environments.

    7. Future of AI: While enterprise AI is moving toward replacing human roles, consumer AI adoption may take longer. The generational shift will likely drive more mainstream consumer use of AI, similar to how personal computing became widespread

• Related to the above: Salesforce Ventures ups its AI fund to $1B, doubling it again

  • At the same time Salesforce dunked on Microsoft copilot it has announced at their massive tech conference, Dreamforce, taking place this week in San Francisco, that its VC arm, Salesforce Ventures will deploy a new $500 million fund dedicated to AI companies

• Cosmos Institue has launched

  • New Institute with the mission to bring philosophy to code, bringing vital concepts around human autonomy, reason and decentralisation into the AI systems that will shape our future

Interesting AI use cases of the week:

• A Chatbot Is Now Cleared to Give Buy, Sell Investment Advice

  • An Israeli regulator has cleared an artificial intelligence startup to launch a chatbot that offers stock-picking advice in partnership with a large bank, even as other governments have raised alarms that AI might destabilize financial markets if used widely in investing. Tel Aviv-based Bridgewise has been given the green light by the Israel Securities Authority (ISA) to release a chatbot called Bridget later this month that can offer recommendations for which stocks to buy and sell in response to user queries. The startup is working with one of the country’s largest banks, Israel Discount Bank, to roll out the product. …The regulator laid down several other conditions for approval, including that the company offering such an AI tool must have an investment license, be compensated on a fixed rate rather than based on how well the investments do, and abide by basic conflict of interest rules

• Eno

  • A groundbreaking generative documentary about visionary musician and artist Brian Eno, a film that’s different every time it’s shown

    • Rich with access to hundreds of hours of never-before-seen footage and unreleased music, Gary Hustwit’s forthcoming documentary Eno employs groundbreaking technology to accomplish something that’s never been done before: a feature film that’s never the same twice. Hustwit and creative technologist Brendan Dawes have developed bespoke generative software designed to sequence scenes and create transitions out of Hustwit’s original interviews with Eno, and Eno’s rich archive of hundreds of hours of never-before-seen footage, and unreleased music. Each screening of Eno is unique, presenting different scenes, order, music, and meant to be experienced live. The generative and infinitely iterative quality of Eno poetically resonates with the artist's own creative practice, his methods of using technology to compose music, and his endless deep dive into the mercurial essence of creativity."

    • According to the film’s makers, there are 52 quintillion possible versions of it! This artistic experiment tells us much about the nature of creativity and the plurality of truth in the age of generative media.

    • Over 500 hours of Eno’s video footage, interviews and recordings were used from an archive spanning 50 years of Eno’s creative output. The filmmakers wrote software generating introductory and concluding scenes with Eno and outlining a rough three-act structure. They then let the software loose on this digital archive, splicing together different scenes and recordings to generate a 90-minute film.

Interesting Start-Up news of the week:

• The end of the 1 billion active user ad-supported consumer startup - and why highly-monetizing, useful, vertical apps are the next thing by Andrew Chen, General Partner for A16Z GAMES

  • Super interesting read as to why it has been such a long time since we last built a broadly horizontal consumer app like YouTube, Instagram, Linkedin, or Snapchat, and why the next generation might be "vertical apps" -- appealing to a vertical segment of the market, like Monopoly Go, Draft Kings, Canva, etc - that are smaller audience products with higher "whale" monetization (sometimes even workflow/B2B)

• PitchBook Analyst Note: Should VC Investors Pick a Lane and Stay in It?

  • Using an AI-powered model, our analysis found that startups whose lead investors had high domain expertise were 1.2x more likely to have a successful exit than those with generalist VCs. Domain sophistication has been especially important at the early stage when a startup is still finding product-market fit and in highly technical areas like pharma and healthtech

• AI start-ups generate money faster than past hyped tech companies

  • New data from payments group Stripe suggest artificial intelligence groups are making revenues at an unprecedented rate

A Start-Up that I like:

• AMT.AI - recently out of EntrepreneurFirst the team have built Lyra - an AI employee for partnering with 1000x more influencers

  • Lyra’s built on the perfect combination of my 4+ years experience doing hundreds of creator partnerships, and Christian’s experience building a scalable adtech platform. Lyra automates all the boring parts of influencer marketing, so you can focus on strategy, not manual work. No more manually checking if tracking links work, no more updating the address on an invoice, no more uploading CSVs to send a mailmerge. Finally, Lyra can do all that work for you.

Science news of the week:

• Making Eggs Without Ovaries by Metacelsus

  • This article discusses a groundbreaking achievement in reproductive biology by a research team at Osaka University, which successfully produced viable eggs from male mouse cells. Building on years of stem cell research, they induced male stem cells to lose their Y chromosome and duplicate the X chromosome, creating female cells from which eggs were generated. The implications for human fertility are significant, particularly in the context of age-related infertility and broader reproductive challenges in the even broader demographic crisis we could face in the not too distant future.

• This Shark Lives 400 Years. Its DNA May Explain Why

  • “These are animals that live longer than human beings, and they do this in the wild, without medicines or hospitals or health care,” said João Pedro de Magalhães, a molecular biogerontologist at the University of Birmingham in England. Studying the sharks, may help scientists one day “develop cancer therapies or prevention measures, or a greater fundamental understanding of cancer that will lead to clinical benefits” in humans.

Sports business news of the week:

• EA Is Launching a New Social App That Mixes Its Sports Game Tech, Ultimate Team, and Live Sports

  • The new app functions similarly to existing sports apps featuring news, scores, stats, and highlights. What is unique is its tie-ins with EA's various sports games, including the possibility of unlocking prizes like Ultimate Team cards based on the sizzle reel.

  • "Our thesis is simple: With the right products, features, and capabilities, a significant portion of hyper-valuable consumer attention is available for a fraction of the cost," Tinson says. EA's goal is to create a social app that allows fans to follow their favourite sports and teams while integrating in-game rewards and community elements.

  • Will this app be able to compete with sports and fantasy apps from ESPN and other media companies?

• Football's TV Rights are under threat

  • Bellingham's new YouTube channel has more subscribers than 7 EPL teams. After 5 days. Long-form media always came after retirement but now it's creeping in during careers. Ronaldo the other week, Bellingham this month - the dominos will keep falling. Fans will crave more BTS than ever before. What does it mean for Football though?

    • Broadcasters must work with Players/Agencies, not against them.

    • Image rights battles will become trickier & more frequent.

    • Matchday can't be the only focus.

    • Fans follow players.

China news of the week:

• China's new plan to fix its broken economy

  • People’s Bank of China has announced: 

    • 800 billion yuan ($113 billion) of liquidity to support Chinese stocks 

    • Plans for a new stock stabilization fund

    • Slashed reserve requirements for banks, short-term interest rates and existing mortgage rates

    • Eased rules for second-home purchases and government purchases of unsold homes

  • The strategy reflects a focus on the supply side of the economic equation, though it doesn’t exactly address weak consumer demand and shaky confidence.

  • China is coming off its worst quarter of growth in five quarters. Much of that lag can be chalked up to a deterioration in its property sector and a painful deflationary spiral.

  • Meanwhile, Beijing’s decision to so explicitly wade into the stock market looks like a bid to attract foreign capital back to China. If we go by the initial market response, the move appears to be working. The onshore CSI 300 Index closed 4.3 percent higher on the day and the offshore Hang Seng Index rose 4 percent — each of their best single-day gains in over two years. 

  • Whether investors stick around long-term is another question.

India news of the week:

• Global firms appoint top India deal makers as Asia heads amid deal surge

  • Firms such as KKR & Co, Blackstone Inc., and more recently Carlyle are expanding the roles of their top executives to oversee the region

  • “Almost all global PE funds have had large India offices operating for over a decade. Given the larger opportunity in India compared to other Asian countries, most global PE funds have chosen India heads to manage the region,” said Bhavin Shah, partner and leader of PE and deals at PwC India.

• Start-up fundraising news:

  • PhysicsWallah, an India-based edtech platform provider, raised over $210 million for its Series B at a $2.8 billion valuation. Investors included Lightspeed, Hornbill Capital and GSV Ventures.

  • Everest Fleet, an India-based ridesharing startup, raised $30 million from an investment by Uber

• India is Home to 6 Visa Temples where Many NRIs Got Visa Boons to Live American Dream or Work in Other Countries

  • At some temples in India, Venkateswara Swamy, Sri Ganesha, and Sankatmochan Hanuman double up as “Visa Gods” for hundreds of thousands of Indians wishing to study, work, live and settle abroad. Pesky paperwork, convoluted process, visa appointment delay, fears of rejection, and likes drive ambitious individuals with their passports in folded hands to the visa temples where the deities are believed to answer their prayers for visa approvals. After all, Indians are known for their steadfast faith in the divine power.

• India Is the Next Great Cheese Frontier, Paneer is old news. A new generation of Indian cheesemakers is wowing global palates with their Gouda and Edam.

  • I like cheese! Can’t wait to try some from India!

UK news of the week:

• 2023 Industrial Electricity prices

• Out of work and unwell: the worrying rise of young people on benefits

  • Rates of disability have risen markedly among people in their twenties and thirties, but seen little change among older groups

    

  • Few countries have seen as sharp an increase in young adult ill health as the UK, where the share of 16- to 24-year-olds reporting a health problem that reduces their ability to carry out day-to-day activities has almost tripled from 7 per cent in 2008 to 20 per cent today.

  • The causes that led us here are still being argued, and there is mounting evidence that some efforts to improve mental health may end up worsening it.

    • This is a very important topic that we need to seriously try and figure out as a society

Academic paper(s) of the week

• The Rapid Adoption of Generative AI

  • "This paper reports results from the first nationally representative U.S. survey of generative AI adoption at work and at home. In August 2024, 39 percent of the U.S. population age 18-64 used generative AI. More than 24 percent of workers used it at least once in the week prior to being surveyed, and nearly one in nine used it every workday. Historical data on usage and mass-market product launches suggest that U.S. adoption of generative AI has been faster than adoption of the personal computer and the internet. Generative AI is a general purpose technology, in the sense that it is used in a wide range of occupations and job tasks at work and at home."

• Are We Adapting to Climate Change?

  • ⁠⁠Using comprehensive panel data across diverse geographies and outcomes, including data on mortality, agricultural productivity, crime, conflict, economic output, and damages from flooding and tropical cyclones, we find limited systematic evidence of adaptation to date. Across 21 outcomes we study, six show a statistically significant declining sensitivity to a changing climate, five show an increasing sensitivity, and the remainder show no statistically significant change. Our results do not imply that specific documented adaptation efforts are ineffective or certain locations have not adapted, but instead that the net effects of existing actions have largely not been successful in meaningfully reducing climate impacts in aggregate. To avoid ongoing and future damages from warming, our results suggest a need to identify promising adaptation strategies and understand how they can be scaled.

    • Important.

• Terrorism Works, for its Supporters

  • Here is a snippet of the abstract but I recommend you read it in full:

    • Under certain conditions, terrorism works, but for its supporters, who compensate the terrorists for their low chance of success and use them as a tool to coerce a government. We conceptualize a terrorist organization as an agent, working at the behest of a base of supporters, who are not themselves members of the terrorist organization, that forms the principal. These supporters provide the resources the terrorist organization needs to carry out its campaign. Even if their own goals are quite moderate, they might still rationally support terrorism, and may even prefer to support terrorists with remarkably extreme goals.

• https://x.com/RiotIQ/status/1837519360933744946

  

  • All writing here direct from RiotIQ:

    • Out of 11 variables (7 cognitive abilities, 3 measures of academic achievement, and general intelligence), long-term memory retrieval peaked at the earliest age (18.1 years), and comprehension-knowledge (i.e., crystallized intelligence) peaked at the latest age (35.6 years).

      

    • General intelligence had a sharp increase in childhood through early adulthood, peaking at age 26.2.

      

    • Fluid intelligence peaked earlier and declined more quickly. Crystallized intelligence peaked much later and declined very slowly. This indicates that learned knowledge lasts much longer into life than the ability to engage in reasoning without context.

      

    • In the images, each line segment represents two test scores for the same person. The thick line represents the average score trajectory at each age, and the two parallel lines around it represent the typical range of scores at different ages. That means there is a lot of variability in cognitive development. Some people peak much earlier or later than the average--and others decline much faster or slower than the average.

Culture news of the week:

• How Did Pop Culture Get So Gloomy?

  • This article explores the shift in pop culture towards darker themes, evident in the rising popularity of horror, dystopian sci-fi, and brooding superhero films. The author highlights how movies and music have become increasingly negative, reflecting a cultural trend toward gloom and dysfunction.

  • It suggests that pop culture trends are deeply interconnected with social dynamics, and while they reflect the audience's mood, they also shape societal emotions and mental health.

  • Here are some interesting charts from the article:

    • Measure of the use of the word “Halloween” over time:

      

    • The increasing use of some supernatural words over the last century

      

      

    • Metrics on the mental health of college students:

• Viral Korean sharpshooter to star as assassin in upcoming series

  • I’m glad culture is working in this case!

• Director James Cameron explains why he is joining Stability AI's board of directors

  • James Cameron: “I’ve spent my career seeking out emerging technologies that push the very boundaries of what’s possible, all in the service of telling incredible stories. I was at the forefront of CGI over three decades ago, and I’ve stayed on the cutting edge since. Now, the intersection of generative AI and CGI image creation is the next wave. The convergence of these two totally different engines of creation will unlock new ways for artists to tell stories in ways we could have never imagined. Stability AI is poised to lead this transformation. I’m delighted to collaborate with Sean, Prem, and the Stability AI team as they shape the future of all visual media.”

Thanks for reading this edition if you made it this far! The next edition coming next Sunday!

1. Top things I’ve read this week

2. AI news of the week

3. Interesting AI use cases of the week

4. A Start-Up that I like

5. China news of the week

6. India news of the week

7. Academic paper(s) of the week

8. Culture of the week

Top things I've read this week:

• Doom Scrolling

  • “We may be close to rediscovering thousands of texts that had been lost for millennia. Their contents may reshape how we understand the Ancient World.”

  • Justin explores the enduring mystery and excitement surrounding the discovery, preservation, and potential rediscovery of lost ancient texts. He highlights the precarious nature of our knowledge of the ancient world, emphasizing that 99% of classical works have been lost. Despite this, recent advancements in technology, like digital unwrapping, have renewed hope for recovering texts, notably at the Villa of the Papyri, which contains thousands of carbonized scrolls from the eruption of Mount Vesuvius. Justin illustrates the transformative impact of such discoveries through examples like Aristotle’s Politics, which, though never lost, was largely unread for centuries. He speculates on the potential rediscovery of long-lost texts, such as Aristotle’s Constitution of the Spartans, and reflects on how such finds could dramatically reshape our understanding of the ancient world.

• The Wrong Kind of City

  • “How much can the way cities grow tell us about the economic trajectory of their countries?”

  • Dietrich Vollrath applies Max Weber’s framework of producer and consumer cities to explore urbanization in developing countries. Producer cities, historically driven by industries and trade, foster innovation and economic dynamism, while consumer cities rely on external rents from resources or privileges, leading to stagnation. Vollrath observes that many rapidly urbanizing developing nations have seen growth in consumer cities, which lack the productivity gains and innovation associated with producer cities. These consumer cities often have inflated costs of living, large slum populations, and lower investment in human capital, limiting long-term growth potential. However, Vollrath suggests that with proper infrastructure and agglomeration, many developing cities could transform into producer cities, driving sustained economic development instead of becoming trapped in the consumer city model.

• Will the EU choose to compete?

  • “Mario Draghi has produced a detailed report that is unsparing in its assessment of the EU's policy mistakes, and surprisingly pragmatic about solutions. Some of them will take time to implement, but in a few cases there's potential for immediate shifts.”

  • The European Union’s challenges in maintaining global economic competitiveness, particularly compared to the U.S. is a topic that has become inescapable. The Diff takes a look at Mario Draghi proposed solutions… Europe’s slower growth trajectory stems from various long-term policy choices, such as prioritizing labor market stability, shorter work hours, and earlier retirements, which hinder GDP growth. Stringent labour protections, tech regulations like GDPR, and fragmented telecom markets have stifled innovation and investment. The article contrasts Europe’s approach to that of the U.S. and China, which foster innovation by attracting talent, investing in strategic industries, and allowing mergers to fuel industry consolidation. The Diff notes that while Draghi's report offers pragmatic solutions, such as streamlining energy policies, easing telecom mergers, and improving defence and space sectors, implementing these reforms remains uncertain, as Europe often struggles to confront economic volatility and competition directly, preferring instead to maintain stability at the cost of growth. We need a shift in the EU's mindset - embrace necessary reforms to enhance competitiveness!

• How to Succeed in a Mr Beast Production

  • Jimmy Donaldson aka MrBeast: "I'm willing to count to one hundred thousand, bury myself alive, or walk a marathon in the world’s largest pairs of shoes if I must. I just want to do what makes me happy and ultimately the viewers happy. This channel is my baby and I've given up my life for it."

  • This is a really interesting insight into 1) How to go viral on Youtube 2) Into the mindset that drives one of the most successful young people in the world. Here are some quotes that stood out to me!

    • "Your goal here is to make the best YOUTUBE videos possible. That’s the number one goal of this production company. It’s not to make the best produced videos. Not to make the funniest videos. Not to make the best looking videos. Not the highest quality videos. It’s to make the best YOUTUBE videos possible."

    • "The Amount of hours you work is irrelevant... you will be judged on results, not hours. We are a results-based company. Get shit done and move the goalpost!"

    • "I only want ‘A Players’… A-Players are obsessive, learn from mistakes, coachable, intelligent, don’t make excuses, believe in YouTube, see the value of this company, and are the best in the goddamn world at their job."

    • "Critical Components... WITHOUT WHAT YOU’RE WORKING ON WE DO NOT HAVE A VIDEO! Protect it at all costs, check in on it 10x a day, obsess over it, make a backup."

    • "Creativity Saves Money... People always assume money is the answer and if we just spend more money we can give Jimmy what he wants. Which is wrong, creativity is the answer."

    • "The three metrics you guys need to care about are Click Thru Rate (CTR), Average View Duration (AVD), and Average View Percentage (AVP)...

  • As an aside: I got a flavour of this mentality when I spent time with Julius Dein, the internet’s biggest magician! I got to watch his deep focus on optimising the virality of Facebook videos in person and it was awesome to see!

• Who migrates from developing countries?

  • “New data from 99 developing countries challenges conventional wisdom in development policy. People whose labour has higher returns at home are more likely to migrate. The model of migration as resembling a costly human capital investment - not a simple occupational choice - is thus more useful for policymakers seeking to shape the real world.” Very interesting!

AI news of the week:

• If you want to deep dive on OpenAI's GPT4o1 then there is no better place to go (as always) then Zvi's substack:

• tldr: Impressive new capabilities centred around Chain of Thought (for general-purpose complex reasoning), mathematics, coding and cryptic crosswords but perhaps not the dramatic step change we are still waiting for

• Also want to call out Ethan Mollick’s take here

  One Useful Thing

  Something New: On OpenAI's "Strawberry" and Reasoning

  I have had access to the much-rumored OpenAI “Strawberry” enhanced reasoning system for awhile, and now that it is public, I can finally share some thoughts. It is amazing, still limited, and, perhaps most importantly, a signal of where things are heading…

  Read more

  2 years ago · 550 likes · 65 comments · Ethan Mollick

  • “o1 lets the AI “think through” a problem before solving it. This lets it address very hard problems that require planning and iteration, like novel math or science questions. In fact, it can now beat human PhD experts in solving extremely hard physics problems. To be clear, o1-preview doesn’t do everything better. It is not a better writer than GPT-4o, for example. But for tasks that require planning, the changes are quite large”

    

• This AI model is helping researchers detect disease based on coughs - Google

  • “Subtle clues hidden within coughs (bioacoustic sounds) hold the potential to revolutionize how we screen, diagnose, monitor and manage a wide range of health conditions like tuberculosis (TB) or chronic obstructive pulmonary disease (COPD).”

  • Google Research's bioacoustic model, HeAR (Health Acoustic Representations), leverages AI to extract health insights from body sounds, such as coughs, using widely accessible smartphone microphones. Trained on 300 million audio samples, including 100 million cough sounds, HeAR excels at recognizing health-related acoustic patterns, outperforming other models and requiring less training data. Indian company Salcit Technologies is using HeAR to improve early detection of tuberculosis (TB) via cough analysis. By enabling low-cost, accessible disease screening, HeAR aims to revolutionize health diagnostics globally, with support from organizations like the UN’s StopTB Partnership.

Interesting AI use cases of the week:

• AI Goose Deterrent

  • AI goose deterrent system - I can't see an option to purchase and it’s not hugely useful for me in London but if you own a jetty and you want to keep geese away then it looks great!

• Not Quite Past It

  • Not Quite Past trains AI to make beautiful things. Their first project explores classic Dutch Delftware tiles which you can design from your own prompts in their Workshop. They then get them made in a kiln in the Staffordshire Potteries and sent to your door!

  • Get ready to see many many more applications like this - long-term I think we should expect to see this as the norm.

• Intelligent Alpha, and First ETF to Deliver Alpha-Generating AI-Powered Investments

  • A chatbot-powered ETF that promises to harness the brainpower of the investment world’s most illustrious minds — Warren Buffett, Stanley Druckenmiller, David Tepper, and more? Sounds great doesn’t it. The Intelligent Livermore ETF, will be built around investment ideas generated by ChatGPT, Gemini and Claude, dubbed the “investment committee,” tasked to emulate the investors’ personalities. The trio of chatbots will spit out 60 to 90 global firms that span a number of sectors, themes and geographies, including health care, renewables and Latin America, to name just a few.

  • What would it mean the best investing robot is an imitation of the best human investors?! I suspect that this won’t be the case but fun nonetheless! Let me know if you’re investing!

• Social AI - replace humans with bots

  • Pick the followers you want e.g. “supporters,” “nerds,” “skeptics,” “visionaries,” and “ideators.” and they’ll engage with you! No human followers allowed. Maybe a good for a trial run before your next tweet?

• Facecam.ai

  • AI faceswap technology that can transform a single photo into a video deepfake - cool technology but you don’t need me to describe the potential risks of this!

Interesting Start-Up news of the week:

• Investors pile into OpenAI’s $6bn funding round in unprecedented bet

  • “Venture groups believe ChatGPT-maker will eventually be the world’s dominant AI company and worth trillions of dollars”

  • Investors such as Thrive and Tiger usually write far smaller cheques for less established start-ups, hoping to make 10 to 100 times their money back. In this case OpenAI, would need to grow to become worth at least $1.5tn! But in the words of the investors “We’re talking about the path to building a trillion-dollar company… I don’t think this is unreasonable.” I would tend to agree, providing of course super-intelligent AI doesn’t end the world before then!

A Start-Up that I like:

• Table.AI

  • “AI meets CRM: Never forget who you've met. Grow and maintain your network effortlessly with AI. Because your network is your net worth.”

  • Check out their video here:

  • Really cool product - I have signed up for access to their beta product once it launches!

China news of the week:

• A Crisis of Confidence Deepens in Xi’s China

  • Details the impact of a nationwide push to lower salaries at state-owned finance companies

  • “President Xi Jinping’s decision to reshape the world’s second-largest economy. Industries such as finance, consumer tech and property — key drivers of China's growth for much of this century — are now out of favour. Instead, the most powerful Communist Party leader since Mao Zedong is funnelling resources toward endeavours such as electric vehicles and chip production. “High quality” growth is the new mantra, not “high speed.””

  • As per Bloomberg “This dislocation poses an existential question: What happens to an economy when you demotivate and demoralize some of your best and brightest workers?” A good question

• New Start-Ups founded in China

  • In 2018, 51,302 new startups were founded. Last year, that number was down to 1,202

  • There has been some debate over the validity of these exact numbers but regardless worth noting!

• Love makes the world go round - Drone Show Guinness World Record

  • In cheerier news check out this 8100 world record drone show in Shenzen - incredible to see

India news of the week:

• The proportion of poorest households that owned a vehicle (motorcycle/scooter,car/jeep)- 2011-12: 6% … 2022-23: 40%!

From <https://x.com/ShamikaRavi/status/1826697208860971176>

Academic paper(s) of the week

• Double-edged stars: Michelin stars, reactivity,and restaurant exits in New York City by Daniel B. Sands

  • ⁠⁠Paper reports that Michelin stars make NYC restaurants more likely to close, due to conflicts they cause with stakeholders, overwhelming the impact of more customers willing to pay more.

Culture:

• I watched the film Civil War -> would highly recommend if only for the concept! Couldn’t not think of one of my favourite films Nightcrawler. To what extent is a war phtographer a passive actor vs an active player? What is the role of people who experience no emotions in society (or at least can successfully suppress them)? I couldn’t help but think Sam Bankman-Fried…

Thanks for reading this first edition if you made it this far! The next edition coming next Sunday!