Category: Thinks

MaxSource: “The fastest-growing companies share a hidden advantage: Every new customer they acquire makes their product more valuable for everyone else. This is network effects in action—one of the most powerful and defensible competitive advantages in business today. Network effects occur when value increases as participation grows. A phone network with two users is nearly useless. A network with ten million users is indispensable. The cost to add new users might stay flat, but the value created compounds. That is why companies built on network effects often dominate entire sectors once they achieve scale.”

Bloomberg: “Because of the way Microsoft is structured, in which its three main product categories—operating systems, productivity software and cloud services—are bundled together, it’s hard to ascribe a precise value to the leading spreadsheet application except to say that without it, there’s zero chance the company that owns it would be worth nearly $4 trillion. In 2025, Microsoft 365 subscription revenue from businesses totaled almost $88 billion, on top of $7 billion from other customers. Those numbers, and Microsoft’s own public disclosures, suggest there are something like 500 million paying Excel users, the rough equivalent of Netflix plus Amazon Prime subscribers. Excel has its corporate challenges, from Google’s web-based knockoff to the looming threat of artificial intelligence, but so far no competitor has managed to mount a serious challenge.”

IEA: “Spontaneous order is crucial for understanding fundamental human institutions (e.g., language and the law, morals, markets and money) and for defending individual liberty. But its operation is often overlooked. Spontaneous orders are self-generating, self-adjusting complex adaptive systems. They exist when a pattern that has not been arranged by any coordinator emerges from the interactions of multiple, dispersed individual elements. Characterised by Adam Smith as the ‘invisible hand’ and by Ferguson and Hayek as ‘the result of human action, but not of human design’, spontaneous order in human institutions is perhaps more clearly understood as the ‘unintended coordination of intentional action’.”

Business Standard: “To stay visible, brands have no choice but to optimise beyond Google’s organic results and win across four interconnected layers — SEO, AEO, GEO, and AIEO. AEO (answer engine optimisation) is designed for visibility in featured snippets, “People Also Ask” boxes, voice search, and knowledge panels. Success in AEO depends on clear, concise, and well-structured content, especially FAQ-style answers, step-by-step guides, and schema markup (FAQ Page, How To …). The goal is to be the source that powers the answer, even if the user never clicks. GEO (generative engine optimisation) is about becoming a trusted citation source for generative AI like ChatGPT, Perplexity, Gemini, and the like. These models retrieve links and synthesise information from authoritative content. To win in GEO, the content should be citation-ready: Actual, well-sourced, recently updated, and backed by real expertise…AIEO (artificial intelligence engine optimisation) focuses on visibility within AI-powered search experiences, such as Google’s AI Overviews and Bing with Copilot.”

WSJ: “A recent report by economists at Goldman Sachs starkly laid this out. In the past, they wrote, 1% more output in China would raise the rest of the world’s output by 0.2% as it pulled in imports. In their new forecast, the Goldman team has concluded that the relationship has turned negative. China’s growth, they write, is being driven by its “leadership’s determination and capability to further advance manufacturing competitiveness and boost exports.” This is positive for other countries insofar as cheaper Chinese goods boost purchasing power. But that benefit is more than offset by the hit to their manufacturing sectors from Chinese competition. The upshot is that Goldman sees China growing 0.5 to 0.8 percentage point a year faster over the next few years, but that will reduce the rest of the world’s growth by 0.1 point a year.”

Carl Frey: “Generative A.I. needs its own course correction — both for the sake of energy efficiency and for its own advancement. Large language models, for all their wonders, can only predict the next thing a human would say. Train one on texts from the late 1800s and it won’t invent airplanes or rockets. It will channel ideas from that period, when leading scientists thought human flight was impossible. If we only scale up our current approach, wasting money on fast-obsolete chips and energy-guzzling data centers, we won’t progress beyond our current technology, which still yields limited, mediocre results…Bubbles are noisy while they inflate. When they burst, the froth clears and you can see which ideas hold up without subsidy. If the A.I. boom cools, what survives will be the systems that do more with less.”

NBC News: “More of Silicon Valley is building on free Chinese AI…AI startups are seeing record valuations, but many are building on a foundation of cheap, free-to-download Chinese AI models.”

NYTimes: “My hope is that I am building a future for [my son] Omie in which A.I. helps create a world where we have two varieties of everything: one that is automated and easily created; the other that is painstakingly made only by humans, and is therefore cherished. I want my son and his generation to still value human-made movies and other art — while also using A.I. to help tell stories about themselves, where they are agile and inventive superheroes, overcoming future challenges we have yet to imagine.”

FT: “Bosses talk endlessly about customers, investors and employees, yet the audience they’re often most anxious to impress is each other. From industry conferences to LinkedIn posts, these are arenas for peer rivalry — who is portrayed as the most visionary, whose message is going viral or whose approach to artificial intelligence or sustainability is best. “Ultimately CEOs are competing against each other,” one headhunter told me. For publicly listed companies, “their primary obligation is to perform better than other CEOs in their sector”. Board directors and investors prioritise relative performance and these executives know it.”

Vasant Dhar on his book “Thinking with Machines”: “I wrote the book because, like it or not, the future is about thinking with machines. And the real question is how we want to shape that future and how we should think about it. The book isn’t about what to think about AI, but how to think about it individually. I also wanted to counter the misconception that AI is recent. It has been around for 60-70 years. As I say in the book, and borrowing from Bob Marley, ‘If you know your history, then you’ll know where you are coming from.’ I wanted to lay out the history of AI, the old problems that remain unsolved in today’s world and the new ones that have emerged around truth, trust and governance.”

WaPo: “Today’s scientists have been trying to explain the puzzling downfall of Harappa, one of the valley’s largest cities, by looking at the environmental conditions. In a study published Thursday in the journal Communications Earth & Environment, an international team used paleoclimate data and computer models to re-create the climate during the civilization’s existence between 3000 and 1000 B.C. They found four intense droughts reduced rainfall and dried up waterways and soils, which probably caused Harappan residents to relocate frequently. “The most surprising finding is that the Harappan decline was driven not by a single catastrophic event, but by repeated, long, and intensifying river droughts lasting centuries,” said Hiren Solanki, lead author at the Indian Institute of Technology at Gandhinagar, India.”

Dwarkesh Patel: “We are moving from the age of podcasts to the age of essays.”

Bloomberg: “Platforms that were engineered to maximize engagement by flooding our feeds may now be driving people away. As news consumption has shifted to social media, avoidance has surged — from 29% in 2017 to 40% in 2025. Young audiences, who live inside these feeds, feel it the most. Many describe the news as toxic and demoralizing, a stream of conflict that leaves them feeling powerless. “The social media era has pushed journalism toward more extreme content,” Newman says. “That’s not what people want.” Research shows that when the body’s fight-or-flight system stays switched on too long, our focus and emotional control begin to fray. News platforms can hijack those same survival circuits, flooding feeds with outrage and alarm until being informed feels indistinguishable from being afraid. Eventually the mind pulls back in self-defense…The unopened alert, the unread terms of service, the avalanche of unread mail silenced with “mark as read”: Avoidance has become our default posture. Ours is the age of the ostrich.”

Ilya Sutskever: “Here’s a perspective that I think might be true. The way ML used to work is that people would just tinker with stuff and try to get interesting results. That’s what’s been going on in the past. Then the scaling insight arrived. Scaling laws, GPT-3, and suddenly everyone realized we should scale. This is an example of how language affects thought. “Scaling” is just one word, but it’s such a powerful word because it informs people what to do. They say, “Let’s try to scale things.” So you say, what are we scaling? Pre-training was the thing to scale. It was a particular scaling recipe. The big breakthrough of pre-training is the realization that this recipe is good. You say, “Hey, if you mix some compute with some data into a neural net of a certain size, you will get results. You will know that you’ll be better if you just scale the recipe up.” This is also great. Companies love this because it gives you a very low-risk way of investing your resources. It’s much harder to invest your resources in research. Compare that. If you research, you need to be like, “Go forth researchers and research and come up with something”, versus get more data, get more compute. You know you’ll get something from pre-training.”

FT: “India has doubled steel production in the past decade, overtaking Japan to become the world’s second-largest steelmaker. Prime Minister Narendra Modi’s government is now aiming to boost India’s steelmaking capacity to 300mn tonnes by 2030, almost double its level in 2024. In a global industry grappling with slowing demand in major markets such as China, where consumption has plateaued, India’s steel use bucks the trend, forecast to rise about 6 per cent a year through 2035. “India is the best story as far as the steel industry is concerned, globally, there’s no other major market growing at this pace — in fact, most markets are shrinking,” says TV Narendran, chief executive at Tata Steel, one of the country’s biggest producers of the metal. “India is underinvested in infrastructure.”…Booming production may buoy the domestic economy, but is causing environmental damage and trade tensions.”

WSJ: “The overall goal of sovereign AI is to build out a localized computing infrastructure and potentially possess domestic capabilities in everything from large language model development and chipmaking to cloud storage and engineering talent. Many also hope to regulate and manage the data used and generated by AI on their own terms…South Korea and others believe their homegrown tech sectors are strong enough to build up their AI capabilities.”

Henry Oliver: “The classical liberal project has become too narrow. It is too often perceived as overly economic and technocratic: as something that ignores the deepest problems of human life, and is unable to speak to the whole human experience. This is wrong! Alongside its advantages for growth and prosperity, liberalism offers greater meaning, dignity, and a sense of community than any of the political alternatives…Liberalism is already gaining new cultural and practical relevance. New interest in the liberal arts is coming from technologists and others online. Classical liberal answers are emerging to solve today’s problems. AI is disrupting the old order of intellectuals, just as the internet once did. The Pursuit of Liberalism is an experiment in this spirit. We are as committed to advocating for human flourishing, as to describing economic reality. We will argue for the enduring relevance of the arts and philosophy to classical liberalism. Our aim is to revive the broad classical liberal tradition. We will argue for its place in an innovative world. We don’t need a new liberalism. We need a new way of helping people to see it for what it is.”

NYTimes: “Think of agentic A.I. as a personal assistant, said Shilpa Ranganathan, the chief product officer at Expedia Group, which is developing both generative and agentic A.I. trip-planning tools. While the more familiar generative A.I. can summarize information and answer questions, agentic tools can carry out tasks. Travelers benefit by deputizing these tools to perform time-consuming chores like tracking flight prices. “Instead of manually running every search or comparison yourself, agentic A.I. can do that work in the background and come back with the most relevant options,” Ms. Ranganathan said. When comparing hotel options, for example, an agentic tool should be able to pinpoint multiple properties’ rates for your requested dates, identify whether they have availability and compile that information in one place so you can easily compare. A generative tool might simply recommend hotels that broadly fit your budget, and offer links to each website, where you can check the details.”

Martin Wolf: “What lies ahead for the world economy? A plausible answer is that it has started to splinter. This is the argument of Neil Shearing, group chief economist at Capital Economics, in his thoughtful new book The Fractured Age. Fracturing, he notes, is not the same as “deglobalising”. Trade and other forms of globalisation might not shrink by much. This should be nothing like the collapse of the 1930s. But trade with rivals will shrink and trade with friends will rise. In particular, he suggests, the world will divide between a US-centred bloc and a China-centred one, with a number of unaligned countries stuck in between, trying to do their best.”

WSJ: “Robots and AI are already remaking the Chinese economy…China’s factories and ports are learning to make and export more goods faster, cheaper and with fewer workers…China installed 295,000 industrial robots last year, nearly nine times as many as the U.S. and more than the rest of the world combined, according to the International Federation of Robotics. China’s stock of operational robots surpassed two million in 2024, the most of any country.”

New Yorker reviews “Capitalism: A Global History”: “[Sven] Beckert identifies “two diametrically opposed stories”: capitalism either deserves credit for the rise in living standards and longevity or stands condemned as an “insatiable demon.” His book addresses “a deep frustration that so many of the stories we tell about capitalism are incomplete and sometimes just plain wrong.” He invites readers to study capitalism “with a sense of wonder, surprise, and astonishment—not because it is ‘good’ or ‘bad’ but because of its world-shaping power, and because understanding it is crucial to navigating our shared future.”” Bloomberg review.

SaaStr: “David Clark at VenCap International (a limited partner in so many of the top VC funds) just analyzed over 1,900 VC funds raised between 2000-2019 using Pitchbook data, and the results are brutal for anyone who thinks generating top-quartile returns is just about “working hard” and “adding value.”…A 5x VC fund is realistic — but only for a small subset of managers who have proven they can access and back the best companies, fund after fund, across cycles. For everyone else, 5x is a pitch deck aspiration that the data says won’t happen. And for LPs, the lesson is simple: your job isn’t to find “potential.” Your job is to find proven access. Because that’s what actually shifts the odds in your favor.”

WSJ: “Riding Amtrak’s California Zephyr route, it turns out, is equal parts sightseeing extravaganza and endurance test…The California Zephyr is one of Amtrak’s most scenic train routes, passing through the Rocky Mountains and Sierra Nevada.” I had done the train journey in 1991.

FT: “As airports in the Middle East expand, they are facing new competition: each other. The region’s hubs in Dubai, Abu Dhabi, Turkey and Saudi Arabia are on an expansion drive that will see them add hundreds of millions of passengers in the coming decade — the equivalent of at least three Heathrows. Dubai, which was once the same size as London’s Gatwick serving around 30mn people a year, plans to increase annual capacity to more than 200mn. Istanbul is also targeting a similar level of passengers, while Saudi Arabia has ambitions to develop its own regional mega-hub through its new airline, Riyadh Air. In the past, the region’s airports eyed Europe’s leading hubs such as Heathrow, Frankfurt and Schiphol as competitors. But as they have grown, executives say they will need to win business from each other.”

WSJ: “People who are immersed in AI often have some surprisingly old-fashioned habits, from scrawling meeting notes on paper to inputting calendar entries manually. Notably, some of the duties that the rest of us are most inclined to give to bots are the very things AI super users insist on doing themselves. This should make us consider whether we are doing AI adoption backward in some cases, or using the latest technology simply because it is shiny and new. AI savants’ greatest strength may be their ability to tell the difference between the to-do’s that robo assistants improve and those that are better done by hand.”

Martin Gurri: “The most relentless enemy of the capitalist system isn’t the proletariat or the revolutionary vanguard but the entitled class.”  [via Arnold Kling]

Bloomberg: “Three years ago, OpenAI released ChatGPT, setting off a mania on Wall Street for all things artificial intelligence. And the stock market hasn’t been the same since. Bets that the groundbreaking technology will reshape society have minted new market leaders, made an already concentrated S&P 500 Index even more top heavy, and left companies and industries considered at risk of being replaced by AI struggling to keep pace…The rally sparked by AI enthusiasm has been the chief driving force behind the S&P 500’s 64% jump since the chatbot’s release. The seven most valuable companies in the S&P 500 — Nvidia Corp., Microsoft Corp., Apple Inc., Alphabet Inc., Amazon.com Inc., Meta Platforms Inc. and Broadcom Inc. — are all major players in the technology and account for nearly half of the benchmark’s gains over that time, according to data compiled by Bloomberg.”

Daren Acemoglu: “Automation isn’t the only thing that AI can do. Machine learning methods and AI more broadly can also help workers, by enabling them to obtain and use better information relevant to their tasks and helping them become more productive and essential for the production process. This type of “pro-worker AI” can support robust wage growth and promote job creation. It would make AI an enabler of shared prosperity, rather than its foe. It is also technically quite feasible. Pro-worker AI is also productivity-enhancing AI. It can help us have much better electricians (with the help of AI tools that enable them to expand their repertoires, understand new electrical equipment and become much better at troubleshooting complex problems), more capable educators, and much better-informed nurses who can ease the bottlenecks in the healthcare system. In each of these areas, and in several others, there are prototype AI technologies that signal the promise of this agenda.”

FT Lunch with Pat Gelsinger: “This opens an opportunity for me to get his snap takes on the other titans of Silicon Valley. Google is “at risk with AI”. For Apple, “time for the next innovation”. Amazon is “building a Google-like position” with chips and the Anthropic partnership. Tesla is bold and brazen, but needs to be more respectful of customers and partners. He pauses at Microsoft. “Essentially, Sam Altman is doing to Microsoft what Bill Gates did to IBM,” Gelsinger eventually says. Gates owned the intellectual property, and made IBM his distribution partner. Microsoft was a relative minnow when the deal was struck in the 1980s, but eclipsed IBM in the 1990s. “The analogy is stunning.””

Pratik Bhadra: “For the past year, marketers have been increasingly focused on utilizing GenAI as a creative assistant. A 2024 report from the American Marketing Association showcases that nearly 90% of professional marketers have used GenAI tools at work. We’ve used it to write email subject lines, generate blog posts and create ad copy. While useful, this approach is painfully incremental. It’s like using a smartphone solely for making calls. The real transformation isn’t using AI as a tool; it’s empowering AI as a teammate. Welcome to “Agentic Marketing.”…Our marketing teams will soon be hybrids of human strategists and AI agents. The brands that win will be those that stop seeing AI as an intern to whom they can delegate small tasks and start treating it as a high-powered co-worker to whom they can delegate entire outcomes.”

Mint: “Major consumer brands—from Coca-Cola to Pidilite and even startups—are increasingly using artificial intelligence (AI) to create quick, low-cost advertisements, launching sharply-targetted campaigns and driving a boom in specialised AI marketing tools. This shift is turning ad production from a resource-heavy human process into a cost-efficient technological utility, allowing companies to quickly test thousands of ad variations across platforms like Instagram, Facebook, and their own websites. This shift is turning ad production from a resource-heavy human process into a cost-efficient technological utility, allowing companies to quickly test thousands of ad variations across platforms like Instagram, Facebook, and their own websites.”

FT: “Once the world’s factory, Beijing’s relentless focus on R&D means the country has become the world’s laboratory.” More also in The Economist: “The country’s high-speed innovation holds lessons for the world.”

The Ken: “Since its launch in India in 2017, digital gold has seen remarkable growth, attracting over 120 million buyers. As of July 2024, more than 40 million people were holding it.”

WSJ: “AI adoption isn’t just a problem afflicting one company, but nearly all of them, according to researchers. The implication is that, as with tech waves past, startups full of keen young people have an opportunity to disrupt incumbents who may be slower to change. Within companies, people who could get the most out of AI might need the most encouragement, education and guidance to get there…Early adopters are piling in, but getting everyone else to use AI in their jobs will take leadership, education and listening to the interns.”

CollabFund: “Starbucks is a coffee company. But in financial terms, it’s also one of the largest banks in America. Customers collectively hold nearly $2 billion in prepaid balances on Starbucks cards and apps—more than 85 percent of U.S. banks hold in total deposits. The twist is that customers didn’t entirely choose this setup. When you pay with the Starbucks app, you can’t simply use your credit card in the moment. You have to preload money into an account. That single design choice—subtle, convenient, and completely intentional—turned every latte purchase into a micro-deposit. Starbucks quietly became a financial institution without ever applying for a banking license. This is how change usually happens in technology: not through rebellion, but through design. What begins as a product feature becomes a regulatory blind spot, and over time, an accepted norm.”

Jimmy Wales: “The defining difference between web 1.0 and the platforms that dominate today is not technological sophistication but moral architecture. Early online communities were transparent about process and purpose. They exposed how information was created, corrected and shared. That visibility generated accountability. People could see how the system worked and participate in fixing its mistakes. Trust emerged not from perfection (there was still plenty of online trolling, flame wars and toxicity), but from openness. Today’s digital landscape reverses that logic. Recommendation algorithms and generative AI models decide what billions of users see, yet their workings remain opaque. When platforms insist their systems are too complex to explain, users are asked to substitute faith for understanding. AI intensifies the problem. Large language models can produce fluent paragraphs and convincing deepfakes. The tools that promised to democratise knowledge now threaten to make knowledge unrecognisable. If everything can be fabricated, the distinction between truth and illusion becomes a matter of persuasion.”

Andrej Karpathy: “AI has been compared to various historical precedents: electricity, industrial revolution, etc., I think the strongest analogy is that of AI as a new computing paradigm (Software 2.0) because both are fundamentally about the automation of digital information processing… Software 1.0 easily automates what you can specify. Software 2.0 easily automates what you can verify.”

FT: “India’s tech development has gone through four main phases, each with its distinctive characteristics: the transformational rise of IT services, the development of software-as-a-service companies, the consumer internet and ecommerce boom, and the emergence of “deep tech” companies focused on complex technologies or hardware…Questions have been asked over India’s level of innovation but the country has created more than 120 unicorns, start-ups valued at more than $1bn, according to Tracxn — the third highest number after the US and China. Indian and international venture capital firms have invested $96bn over the past five years, according to consultants Bain, in around 8,000 funding rounds. Most of this has come from foreign investors but the domestic long-term capital base is developing fast, from family offices to insurance companies and pension funds. That should help spur the new generations of hungry Indian entrepreneurs, inspired by the Silicon Valley playbook.”

WSJ: “How often have we heard it: Stay busy to make the most of the time we have left. But there’s a lot to be said for doing the opposite.”

Business Standard: “The logistics cost in India for decades was treated as a dysfunctional economic reality, estimated at a staggering 13-14 per cent of gross domestic product (GDP), far above that of its global peers. This aspect was often cited as the “hidden tax”, which made Indian manufacturing uncompetitive. On September 20, that belief was quietly but decisively shattered. The government, after a detailed nationwide study by the National Council of Applied Economic Research (NCAER) and Department for Promotion of Industry and Internal Trade (DPIIT), released the first credible, data-driven estimate of logistics cost at 7.97 per cent of GDP for 2023-24 (FY24)….India’s logistics story is no longer about a crippling disadvantage. It is about an economy in transition — aligning with global peers, reducing inefficiencies, and building the backbone for a $5 trillion economy. The debate must shift from exaggerated handicaps to ambitious possibilities.”

Benedict Evans: AI Eats the World

FT: “My conclusion is this: getting to a general-purpose quantum computer — the kind that works like a normal computer but has the exponential processing power of quantum mechanics able to explore a vast number of possibilities at once — requires upwards of 1mn physical qubits (quantum bits). This needs a technological leap of equal magnitude…It is time for the superconducting qubit community to shift its focus from chasing the next algorithmic demonstration to tackling the immense manufacturing and engineering challenge that lies ahead. The moment for foundational scientific discovery needs to give way to the era of industrial manufacturing. We have much of the physics; now we need the engineers and technicians. Let us bring the needed manufacturing technology to bear and make this happen quickly, or we risk letting the potential of quantum computing remain forever trapped in a jungle of wires.”

WSJ: “The key thing to remember is that—much like the AI boom—the GLP-1 surge is still in its infancy. Lilly only began selling its weight-loss drug Zepbound in late 2023, and the Food and Drug Administration only declared an end to a supply shortage of obesity drugs last year. As production has scaled up and new clinical data has emerged, Zepbound has pulled ahead of Wegovy. Despite Zepbound’s later launch, Lilly now captures a clear majority of new obesity-drug prescriptions, a sharp shift in market dynamic.”

Chris Gilliard explains “luxury surveillance”: “I argue that an ankle monitor and an Apple Watch are essentially the same technology. They’re essentially the same thing. And that we can really understand some of the things that tech companies are up to when we tease out those parallels. That there’s a segment of devices often chosen by people who have the ability to say yes to surveillance where they think that they are likely to gain some benefits and they don’t understand them as surveillance devices. And that ultimately a big part of this is helping to normalize surveillance for all of us by the fact that people embrace these technologies and understand them as aspirational.”