A careful study of every populist episode since 1900 finds catastrophic consequences, which play out slowly. On average, incomes fall behind by nearly 15% over 15 years. For the U.S., this is a cost of about $13k per person per year. Over a lifetime, that's million bucks.

There’s a deep malaise in tech right now. —New grads can’t find jobs —BigTech middle managers are trying to justify their existence —Everyone not in AI wants to be in AI —Founders struggling with their startup for years see Roy rewrite the rules —Comp insecurity is at an all time high with the Meta offers (“why am I working so hard?”) Tech, net of AI, is just not as sexy a job it used to be 10yrs ago. Tech, net of AI, is not the sexy job it used to be.

I am unreasonably excited about self-driving. It will be the first technology in many decades to visibly terraform outdoor physical spaces and way of life. Less parked cars. Less parking lots. Much greater safety for people in and out of cars. Less noise pollution. More space reclaimed for humans. Human brain cycles and attention capital freed up from “lane following” to other pursuits. Cheaper, faster, programmable delivery of physical items and goods. It won’t happen overnight but there will be the era before and the era after.

"'Artificial intelligence is going to replace literally half of all white-collar workers in the U.S.,' Ford Motor Chief Executive Jim Farley said in an interview last week w/ author Walter Isaacson at the Aspen Ideas Festival. 'AI will leave a lot of white-collar people behind.'”

Cloudflare CEO @eastdakota is having the most honest conversations I've come across about the current & future of content creation "6 months ago, 75% of queries to Google get answered on Google. Which means if you're an original content creator, your content is getting summarized & sold (they still put ads there), but you don't get that traffic. And that's the good news. It used to be that for every 2 pages G scraped, you would expect 1 visitor. 6 months ago that deteriorated to 6 pages scraped to get 1 visitor. Today the traffic ratio is: for every 18 pages Google scrapes, you get 1 visitor. What changed? AI Overviews If the business model of the web has been search, fundamentally, for the last 35 years. You get value by subscriptions, ads, or fame. All 3 of those things are going away, and they are going away fast. And that's STILL the good news. What's the ration for OpenAI? 6 months ago it was 250:1. Today it's 1,500:1. What's changed? People trust the AI more, so they're not reading original content. People aren't following the footnotes. So if you believe the business model of original content creation is driving people to that content... I just have a really bad story for you. The future of the web is going to be people reading the summaries of content, not the original. What I'm worried about is - if you can't sell subscriptions or monetize ads or get the ego boost from people reading your stuff, why anyone is going to create content?"

OK, this is an original and good point.

A few people have asked me if a technical fix for AI model sycophancy is on the cards. In fact, a technical fix for sycophancy is trivial. In many cases all it would take is a tweak to the system prompt. The reason companies are struggling to get this right is not technical. It's because most of us want our models to flatter us, even though we'd never admit it and we all think we're the exception. And it has to be subtle, otherwise we feel insulted by being reminded that we like sycophants. Also, some kinds of sycophancy can have safety implications, such as when the user is struggling with mental health. If a model developer eliminated sycophancy, their scores on arena evaluations would tank and users would perceive the model as less helpful. So they need to find the sweet spot. And the optimal balance itself changes over time because of users' gradually shifting expectations of how models should behave — expectations that are in turn shaped by how models' behavior has been shifting over time. So developers have to solve a game theory problem. Chatbots have been in widespread use for almost 3 years now; it would be nice to be able to track how their behavior on various dimensions (including sycophancy) has been changing over time.

My least favorite thing about SF is how people are constantly sizing you up see how much they can gain by becoming your “friend”. Assessing who you know, how you got into a party, where you’re from or where you live, as basis for whether or not you are worth even speaking to.

LLMs do more than pattern recognition and less than reasoning. We don’t have a name for it yet.

Solar in Pakistan is one example of the global solar boom and why petrostate politicos are freaking out. Solar is doing more to address energy poverty than any other energy source. High costs of fossil fuels still cripple poor nations. e360.yale.edu/features/pakis…

As a neurosurgeon I care a lot about road safety. By now you’ve probably seen @Waymo’s stunning safety results (like 91% fewer serious crashes). But they didn’t just publish data headlines. They released the raw CSV files and data dictionaries. I did a much deeper analysis. A fascinating story emerges when you analyze how they’re achieving this. This isn’t incremental improvement - it’s categorical. We’re looking at the potential elimination of traffic deaths as a leading cause of mortality. The intersection breakthrough: Waymo has essentially solved intersection crashes, with 95% fewer injury incidents than human drivers in the same locations. That’s transforming the deadliest driving scenario. The national math: If every US vehicle performed like Waymo, we’d prevent 33,000-39,000 deaths annually and save $0.9-1.25 trillion in societal costs. Even partial adoption at 27% would save ~10,000 lives per year. In terms of magnitude, this would be the equivalent of eliminating every pedestrian death nationally in a year. The physics signature: Here’s what fascinates me: 47% of Waymo’s contacts involve less than 1 mph delta-V. They’re not just avoiding crashes; they’re converting unavoidable incidents into gentle bumps. It’s like having physics itself on your side. We’re not talking about marginal safety gains. The data represents a fundamental shift from harm reduction to harm prevention. The methodology matters: I used their dynamic geographic benchmarks (comparing like-for-like road conditions) and verified the findings hold across San Francisco, Phoenix, LA, and Austin. The safety advantage actually increases in more complex urban environments. Link to raw data below…. Notes on my approach: Analysis based on 96 million miles of Waymo Rider-Only (RO) data through June 2025, utilizing Waymo's dynamic geographic benchmarks to compare Waymo Driver performance against human drivers under similar road conditions and operational design domains. The projections for national impact (deaths prevented, societal costs) involve several assumptions. Given Waymo's zero reported fatalities, the direct serious injury reductions were mapped to national fatality statistics using established NHTSA-derived ratios that correlate serious injury crash rates with fatality rates. This extrapolation assumes that Waymo's observed serious injury prevention capability would translate proportionally to fatality prevention. Societal cost savings are estimated by applying average per-fatality and per-injury economic costs (e.g., medical, lost productivity, quality of life) as published by NHTSA, scaling these national averages to the projected number of avoided fatalities and injuries based on Waymo's safety performance. These figures represent the potential annual impact if the Waymo Driver's safety profile were widely integrated into the national fleet. @ethanteicher

"we're selling shares of private companies that don't actually represent any real equity or ownership to retail investors...they don't know the difference"

What was behind biggest-ever U.S. honey bee die-off? Hint: It wasn't pesticides. The U.S. just lost 62% of its commercial honey bee colonies in 7 months. Here’s what happened - and why it should terrify anyone who eats. 👇1/

My baseline case is that the AI being built right now is overrated; soon it will be a disappointment; then it will be a bubble; and, by the 2030s, it will be world-changing. Self-driving cars are a model for this. - In 2015, I heard autonomy was 5 years away from taking over the roads. - In 2020, they were nowhere. - Even in 2022, you could say they were a huge disappointment. - Now, they're quietly a revolution. Driverless taxi usage in CA grew 8x in one year, and Waymo is expanding to other cities.

Breakneck: China's Quest to Engineer the Future, out August 26 danwang.co/breakneck/

Fascinating look at how late medieval England turned itself into a workers' hell by passing maximum-wage laws(!), blocking people from getting better jobs, moving to other villages, or even simply shifting from what they were doing as 12-yr-olds. ageofinvention.xyz/p/age-of-…

It was a busy week! Here’s a roundup of what we launched: — An advanced version of Gemini 2.5 Deep Think achieved gold-medal level performance at the 2025 ICPC World Finals — You can now share your @GeminiApp Gems with anyone — At this year’s #MadeonYouTube event, we introduced Veo 3 Fast and additional Veo controls coming soon to @Youtube — You can now access Gemini in @GoogleChrome directly from your browser — We announced the Agent Payments Protocol (AP2) to securely enable agent-led transactions. Now, users, merchants, and payment providers can transact with confidence across all payment methods

You may have seen reports that say the negotiations have gotten tense — and one from the WSJ recently claims that OpenAI believes they're close to declaring they've reached AGI. A source familiar with the matter confirms, and breaks it down further:

The economy is on the precipice of recession. That’s the clear takeaway from last week’s economic data dump. Consumer spending has flatlined, construction and manufacturing are contracting, and employment is set to fall. And with inflation on the rise, it is tough for the Fed to come to the rescue.

Something new from me 👇

Most people assume China already sees itself as the world’s top manufacturing power. But that’s not how China sees it. China sees itself as the world’s factory, which is not the same — and according to one of the most important official reports used by Chinese policymakers, the United States is still ranked number one in actual manufacturing strength overall. This report, "China Manufacturing Power Development Index Report," published annually for over 15 years now by the Chinese Academy of Engineering, the China Center for Information Industry Development, and CCID Consulting, evaluates countries not just on how much they produce, but on the full range of capabilities that make a country a real industrial powerhouse. It’s one of the reference points that policymakers care about deeply when assessing China’s progress and setting long-term priorities. I recently read the full 2021 report (which ranked countries as of 2020). As of that year, the United States was firmly in the lead, scoring 173.19. China was still in fourth place, behind Germany and Japan. The U.S. had also gained the most in absolute score since 2015. At that time (2021), the index had four major dimensions, each with meaningful weight: 1. Quality and efficiency (29.31%) – Captures labor productivity, value-added rate, profitability, and brand strength. 2. Structural optimization (28.05%) – Reflects the technological depth and composition of the manufacturing sector, including R&D intensity and patent power. 3. Sustainable development (23.13%) – Measures energy efficiency, environmental impact, and digital transformation. 4. Scale development (19.51%) – Covers total manufacturing output and global export share, but is the least important of the four. So it’s not about who produces the most stuff. It’s about who does it best — with the most advanced systems, the most efficient use of resources, and the strongest long-term competitiveness. I couldn't find the full version of the 2024 news report, but it seems that since 2022, the index has now added a fifth dimension: innovation development. I don't have the new weights, but we do know this: the U.S. is still number one by a clear margin, and China remains at number four. You can see that in the screen cap below. So why is the US still #1 in China’s eyes? Because actually ... despite no longer being a maker of many things, the US still leads across nearly every major dimension. China, by contrast, scores well on scale and has made visible progress in structural optimization and digitalization, but it still trails in key areas like energy efficiency, profitability, and global brand recognition. The deeper takeaway is this: when people talk about China surpassing the US in manufacturing, they’re usually thinking about where goods are made. But from the perspective of China’s own top engineers and planners, being a manufacturing superpower is about much more than that. It’s about upstream capabilities, technological sophistication, and resilience over time.

My colleague @skesslr and I put a lot of time and care into this heartbreaking story of a man whose delusional conversations with ChatGPT ended in his suicide and the murder of his mother. wsj.com/tech/ai/chatgpt-ai-s… via @WSJ

New newsletter: Political violence is like a lightning bolt: sudden, seemingly random, but emerging from a local weather system. Today's post is about one particular modern weather system that I think we need to talk about in relation to this, and other, acts of political violence, which is the phenomenon of terminally online young adulthood, especially as it relates to young men. 4 facts, each with screenshot: 1. ATUS data shows that young men, and single men without kids, have seen the largest increase in solitary leisure time in the last 20 years. 2. Work by @jayvanbavel and his lab have shown that the Internet isn't just the physical world on glass. It's a completely different conversation environment, where negativity and extreme opinions drive engagement while out-group animosity and "moral-emotional" language go viral. 3. Work by Michael Bang Petersen et al have shown that "need for chaos" voters—the most nihilistic sliver of the electorate—tend to have the highest levels of social isolation 4. The FBI and other law enforcement agencies around the world have said that political violence is becoming less group-based and ideological and more lone wolf with "salad-bar extremism," which is making it harder to trace political violence before it hits These facts all fit together. As young single men have dramatically increased their time alone and online, they’ve marinated in a unique attentional environment that is more charged with extremist ideas and emotional negativity. Political scientists have found that social isolation increases the risk that young men develop a “need for chaos,” and law enforcement officers have independently confirmed that modern political violence is more likely the result of isolated lone wolves who stitch together a bespoke ideology of hatred that is disconnected from any formal organization. derekthompson.org/p/all-the-…

Mike From PA asks the author of Abundance if he’s read Abundance. Absolutely genius.

someone who's leaving xai told me they decided to "choose life" ❤️

After years of delays, the fastest train in the U.S. is finally here. All aboard @Amtrak’s next-gen Acela debuting Thursday across the Northeast Corridor between Boston and D.C. and stops in between. I just rode the train from NYC to Boston. Here’s what you need to know. (1/7)

To understand the twists and turns that the Epic-Google case took to get to this point today, I recommend my new book that explores this deeply. "iWar" is about Apple but also about the battles @TimSweeneyEpic faced against Google.

DeepMind founder Demis Hassabis on how he set up his life to maximize deep thinking with two full workdays in a 24-hour period:

The new Gemini 2.5 image model🍌is by far the best out there with a whopping +180 ELO point lead in image editing & it really excels at character consistency. Available for free in the @GeminiApp right now. Try uploading an image & playing around with it, it's pretty amazing!