Category: AI

Categories
AI Podcasts

A Remarkable Conversation…

Highly recommend this conversation between Harry Stebbings and Clay Bavor. Among many topics, I especially enjoyed the discussion about not investing in frontier models, the important values, the particular importance of craftsmanship, intensity, and family. And the special conversation about parenting and kids near the end. Just a delightful conversation to be able to enjoy!

Key Highlights:

• Founding Sierra: Bavor explains why he and Taylor chose to start Sierra, focusing on the transformative potential of language model-based agents (1:37 – 5:53).
• The AI Tech Stack: Sierra focuses on building enterprise-grade agent architectures and fine-tuning models on top of open-weights models rather than pre-training foundation models from scratch, prioritizing capital efficiency (5:53 – 7:15).
• Unbounded Demand for Intelligence: Bavor argues that there is massive, unmet demand for “frontier-level” intelligence in fields like coding, science, and legal work (7:15 – 11:41).
• Internal AI Operations: He details the use of Pinecone, an internal AI agent Sierra developed to navigate company data, streamline engineering, and assist in recruitment (18:36 – 22:00).
• Enterprise Strategy: Sierra employs a “forward-deployed” engineering model, embedding staff within client companies to ensure rapid, effective integration of AI, leading to quick deployment timelines (30:12 – 33:22).
• Board Governance: To keep pace with the speed of AI development, Sierra operates on a six-week board meeting cadence, utilizing comprehensive memos instead of traditional slide decks (39:07 – 41:13).
• Corporate Culture: Bavor emphasizes values like craftsmanship, intensity, and family. He also highlights the importance of working in-person to foster apprenticeship, mentorship, and a cohesive team culture (43:02 – 55:41).

Categories
AI AI: Large Language Models Apple

The Slipstream Strategy

Apple had a problem no amount of money could solve. An iPhone can’t draw the power or shed the heat of a data center, so ten different tasks can’t mean ten different models fighting for the same sliver of RAM. Apple’s answer was to freeze one small, efficient base model into the device and then swap tiny adapters in and out of it in milliseconds — a summarization adapter for your texts, a Siri adapter for on-screen actions, and a handoff to Private Cloud Compute for anything heavier. The phone behaves like it’s running many models. It’s running one model wearing many hats.

That architecture — a frozen base plus swappable adapters — is quietly becoming the default way serious AI companies build, and it’s worth understanding why, because it inverts the assumption most people still carry into this industry.

The assumption is that winning means owning a frontier model. Sierra co-founder Clay Bavor pushed back on that on a recent 20VC episode: pouring capital into your own pre-training, he argued, tends to leave you holding a highly perishable bag of floating-point numbers. Open-weight models improve fast enough that yesterday’s frontier is next quarter’s commodity. The companies playing this well aren’t racing to out-spend the labs. They’re slipstreaming behind them — taking the free, state-of-the-art engine and putting all their effort into what sits on top of it.

What sits on top is LoRA — low-rank adaptation. The old failure mode was catastrophic forgetting: fine-tune a model hard enough on your own data and it forgets how to reason generally. LoRA sidesteps this by leaving the base model untouched and training a small set of additional parameters alongside it — a thin layer of expertise bolted onto a frozen foundation. You get real domain depth without touching the thing that makes the model work at all.

The business logic that follows from this is the actual point, and it’s simpler than it looks:

You stop being hostage to any one model provider — if a better open-weight model ships next month, you port your adapter, not your whole product. You can serve hundreds of differently-customized clients off one base model on one piece of hardware, instead of running a separate giant model per customer. You can ship a fix in an afternoon, because an adapter is a few hundred megabytes, not a training run. And in regulated industries, your proprietary data can train an adapter that never leaves your own infrastructure.

None of this is really a story about model architecture. It’s a story about where the moat moved. For a while the moat was raw capability — whoever had the best model won. Apple and Sierra are betting the moat is now somewhere else entirely: in how tightly you can weave a commodity intelligence into a specific workflow, a specific dataset, a specific customer relationship. The engine is free. The adapter is the business.

Categories
AI Apple Google

The Floor

I compared the frontier to a three-star chef making grilled cheese in “Context Rot” — the smartest models on earth spending most of their time on work beneath them, the way a chef trained at Le Bernardin might still melt cheese between two slices of bread on a Tuesday night and call it dinner. The comfort was the point: if the sharpest tool is saved for hard problems and something merely-very-good handles the rest, nobody’s losing anything. The floor was never the interesting part.

I’ve kept turning the joke over, and I think I had the wrong worry.

Watch what companies do with their AI spend, not what they say. Coinbase moved engineers off frontier models onto open weights and cut its AI spend nearly in half while usage kept climbing. Nvidia runs a closed model as orchestrator and routes the actual volume — the daily uncelebrated bulk of it — to open weights it controls. The frontier is becoming a dispatcher, deciding where the request goes and rarely doing the work itself. The instinct is to worry about whose open weights end up running that volume, and right now the most capable ones at scale are Chinese — GLM, Kimi — which makes it tempting to read this as a contest America is quietly losing: the floor of the AI economy built somewhere else, at a price export controls can’t touch. You cannot embargo a file already downloaded. You cannot price-match free.

But that framing has a hole. Google’s own Gemma family is open-weight and good enough to handle that daily volume without anyone reaching for GLM or Kimi. “Open weights are a Chinese story” only holds if you don’t count the open models the company running Android and half the internet’s search traffic has already shipped.

And once I saw that hole, a bigger one opened behind it. I’ve been trying Apple’s new Siri — arriving with iOS 27 this fall, genuinely surprisingly good in beta — and it made me realize open weights, of any nationality, were never going to cook most of the world’s dinners. Apple and Google are.

Consider what actually determines where the world’s routine inference runs. Not which model benchmarks best, not which weights are downloadable — what’s already installed. Apple ships to well over a billion active devices before routing a single query through Siri’s new architecture. Nobody has to be persuaded to try it, or hear about it on a podcast; it’s the thing that answers when you press the button you’ve pressed for a decade. Google owns the search bar and the Android default the same way. Between them, that’s most of the world’s phones — and phones are where most of the world’s questions get asked.

The open-weight framing assumes the floor is up for grabs, that whoever ships the best free model wins the daily grind by merit. But the floor was never a bazaar. It’s a set of defaults, owned by whoever already has the device in your hand, not whoever holds the most generous license. Apple didn’t need to win the model war to win this. Its heaviest reasoning tier is built with Google, running on Nvidia chips in Google’s cloud, under a deal reported at roughly a billion dollars a year — Apple doesn’t fully own the engine doing the thinking. It doesn’t need to. It owns the button.

That’s a quieter concentration than an export-controls fight, and a harder one to dislodge. An open model can be forked, distilled, undercut, or out-competed by the next release. A billion phones with an assistant built into the lock screen cannot be routed around. Whoever’s weights hum underneath barely matters, the way it barely matters to a diner which supplier delivered the flour. What matters is whose kitchen the meal came from, and whose name is on the door.

The grilled-cheese chef was never the risk. Two chefs are about to own nearly every kitchen on earth, and most of us will never notice — because a kitchen you’ve been eating out of for a decade doesn’t feel like something that was won. It just feels like home.

Owning the kitchen and getting paid for what’s cooked in it, though, turn out to be two different questions. That one’s for another post.

Categories
AI China Youth

The Arithmetic of Youth

The first meeting was at one of the banks on a high floor somewhere in Shanghai, the kind of view that turns a city into an abstraction. It was 2005, and I was there the way American investors were there that year — curious, a little jet-lagged, trying to read a country that was rewriting itself faster than anyone could print the new edition. Across the table sat a management team, and what struck me wasn’t anything they said. It was how young they were. Not junior-young. Running-the-company young.

Afterward — in the hallway or the car, early in the trip, when I still had the confidence of someone who thought he could just ask — I put the question to one of our local colleagues. Casually, expecting a casual answer. Something about a young country, a young economy, energy meeting opportunity.

The answer I got instead was the Cultural Revolution.

There was a generation, she explained, that simply wasn’t there. Sent to the countryside, pulled out of universities, handed shovels instead of textbooks. By the time China opened back up, that cohort had a hole in it — a rung missing from the ladder. So the young people I’d just watched run that meeting weren’t there because anyone had bet on youth. They were there because there was no one older left to put in the chair. Youth, in that boardroom, wasn’t a strategy. It was a vacancy dressed up as one.

I have thought about that answer, off and on, for twenty years, without knowing what to do with it. Then a few weeks ago I read a summary of a conversation with Nathan Lambert — an AI researcher who’d just spent time visiting the frontier labs in Beijing and Hangzhou — and I found myself back in that room, except everything about the youth in it had flipped.

He describes teams at places like Moonshot AI as almost absurdly young, tight-knit, close to giddy about the work — “the best vibes,” he calls it. Zhipu AI, he says, has built something close to an AGI showroom, a physical space engineered to perform confidence for whoever walks through the door. These aren’t companies with a hole where the experienced people should be. These are companies that went looking for twenty-five-year-olds because twenty-five-year-olds move at the speed frontier AI research demands, and installed them at the center of the room. The showroom isn’t hiding a vacancy. It’s staging a choice. That’s panel two.

Same demographic. Same first city — Beijing both times — with a high-speed rail line now running to Hangzhou instead of whatever second city I’d have named twenty years ago. Opposite cause. In 2005, youth in the room meant a generation had been taken from the labor force involuntarily. In 2026, youth in the room means a generation has been selected for it, deliberately, competitively, because being young is now the qualification rather than the disqualifier. The Cultural Revolution left a gap that youth filled by default. The AI boom left a door that youth is filling by design.

I would have stopped there, satisfied with the irony, except for a number I couldn’t get out of my head once I went looking for it: 15.6 percent. That’s China’s urban youth unemployment rate — ages sixteen to twenty-four, university students excluded — as of May 2026, and it counts as good news, down from 16.3 percent in April. A year earlier it had spiked to nearly nineteen percent in a single August, the month twelve million university graduates walked out of commencement and into a labor market that had no idea what to do with them. Some will sit for civil service exams, chasing what people there still call the iron rice bowl — the illusion of permanence a state job used to guarantee, back when your grandparents didn’t choose their careers so much as get assigned them. Others will enroll in another degree, not because they want one, but because a classroom is a more dignified place to wait than an unemployment line.

So there is a third panel now, and it doesn’t fit neatly next to the other two. It isn’t a vacancy, and it isn’t a showroom. It’s just a very large number of young Chinese people who did everything they were told to do — studied hard, got the degree — and are standing outside a door that isn’t opening. And somewhere behind that door, in a much smaller room with much better lighting, another group of young Chinese people, maybe the same graduating class, are building the technology that a Silicon Valley researcher travels overseas to admire for its vibes.

I don’t think those two rooms are as separate as they look. I think the showroom is real, and I think the twelve million are real, and I think the mistake — my mistake, sitting here in Menlo Park two decades removed from that conference table — is letting either one stand in for “Chinese youth” as if it were a single sentence instead of a population. The Moonshot AI team is not a representative sample. It’s the visible sliver of a generation, selected with a precision that turns the unemployment numbers into part of the same mechanism — one sorting process, not two unrelated stories. The best vibes in that lab and the worst numbers in that economy might just be describing the two ends of the same funnel.

I keep coming back to that hallway in 2005, and to how confident I was in the question I asked — as if a generation’s youth could only ever be telling one story. It couldn’t then, and it can’t now. I got a true answer that day and thought I understood something. I understood one panel of a triptych I hadn’t seen the rest of yet — and I’m still not sure I’ve seen all of it.

Categories
AI

Context Rot

Here is a small, possibly embarrassing confession: I have never, not once, gone looking for the best AI model.

I have a model. It lives in a browser tab — Safari, usually, on whichever device is nearest, occasionally Chrome if I happen to be at the desktop. It does what I need — drafts an email, untangles a sentence, tells me what a Norwegian emigration record from 1856 probably says — and then I close the tab and go on a walk.

Somewhere out there, presumably, a much smarter, much more expensive machine is doing something extraordinary with protein folding or hedge fund arbitrage or the outer edges of mathematics I will never visit. I have made my peace with never meeting it.

This did not used to feel like a confession. For a while there — a year, eighteen months — it felt like the central drama of the whole industry: which model was “best,” who had it, who had lost it, whether some lab’s quarterly earnings call would reveal that the frontier had quietly moved sixty miles down the road while everyone was looking the other way. Benchmarks were released like box scores. People argued about them the way people argue about batting averages, with the same weird intensity, the same conviction that a two-point difference in some abstract reasoning test settled something important about the future.

And then, at some point I can’t quite date — it crept up, the way these things do — I noticed I had stopped caring.

Not because the frontier stopped moving. It didn’t. It’s still moving, arguably faster than ever, in ways that occasionally show up in the news with all the drama of a soap opera (a delayed launch, a researcher poached, a stock down five percent in an afternoon, always something).

I stopped caring because none of it touched me. My model — whatever it was, this week — had long since crossed some invisible threshold past which more didn’t register as more. It was already better than I needed. It has been better than I needed for a while now. I suspect I am not unusual in this. I suspect most people, doing most things, most days, are operating comfortably inside a capability surplus so large they’ve stopped noticing it’s there, the way you stop noticing a room is warm.

If the top of the model isn’t for people like me — and it increasingly isn’t — then who, or what, is it actually for? I went looking for one piece of the answer and found, instead, a metaphor.

It’s called “context rot.” I have to admit, before I go further, that I’m not sure I’ve ever felt it myself — which, on reflection, is its own small piece of evidence. My sessions close in minutes, not hours. I ask, it answers, I leave. Whatever happens to a model over the fourth or fifth hour of sustained, dependent work is a country I simply don’t visit.

But other people do, increasingly — entire teams do, for entire projects — and what they’re finding out there is worth understanding, even secondhand. It describes something that happens to AI models when they’re asked to work for a long time on something complicated — not five minutes, but five hours; not one question, but a hundred small decisions stacked on top of each other, each one depending on the last.

You’d think the limiting factor would be room. Models have a “context window” — a stated capacity, like a gas tank, measured in tokens, and for a while the marketing numbers on these were the whole story: two million tokens! A library! And you’d think, as with a gas tank, that the thing runs fine until it’s empty and then it stops.

That is not, it turns out, what happens. What happens is closer to what happens to your desk.

You know the desk. Everyone has the desk. It starts the morning clean — an aspirational, almost insulting cleanliness — and by four in the afternoon it is a geological record of the day: three coffee cups, a stack of things you meant to file, a Post-it with a phone number you no longer need, the good pen buried under a printout of something you already dealt with an hour ago. The desk is not full. There is, technically, room. You could clear a space if you tried. But you don’t try, because functionally, cognitively, the desk has stopped being usable long before it ran out of surface area. You start looking for the stapler and forget what you were stapling. This — and I did not make this term up, I want to be clear, though I wish I had — is context rot. The window hasn’t run out. The signal has just drowned in its own debris.

Researchers watching this happen to long-running AI agents have found something almost cruelly elegant about how it fails: it doesn’t fail gradually, the way you’d expect a desk to get gradually messier. Errors compound. A task that takes twice as long doesn’t get twice as likely to go wrong — the failure rate roughly quadruples. Two mistakes early in a long chain of dependent steps don’t add up to a slightly worse outcome. They multiply into something close to total collapse, four hours in, for reasons that trace back to a single bad assumption made in the first twenty minutes and never revisited.

Here is where the frontier comes back in — not as the whole answer, but as a piece of one.

It is not that frontier models are smarter in the way a benchmark measures smart — better at a single hard math problem, a cleverer turn of reasoning. Plenty of models can do that now; the “good enough” tier has crept remarkably high.

It’s that frontier models are apparently, marginally, meaningfully better at not rotting. At keeping the desk usable at hour six. At knowing which of the forty things on the desk actually still matters and which is a coffee cup that should have been thrown out an hour ago. This is a genuinely different kind of intelligence than the one benchmarks were built to measure, and it is almost invisible from the outside — you don’t see it in a single exchange, you see it only in the difference between a project that holds together over three days and one that quietly, subtly, stops making sense somewhere around Tuesday afternoon and nobody notices until Thursday.

If that’s true — if the frontier’s real edge is durability rather than raw cleverness — you’d expect to see it show up in how the labs actually deploy their own models: saving the sharpest tools for the tasks that need to survive the longest.

I went looking for a real-world example and found one closer to home than I expected: Anthropic’s own Slack tool, the one where you tag the AI into a channel the way you’d tag a coworker, and it works alongside a whole team over days, learning the channel as it goes. It runs on a serious, capable, thoroughly frontier model — but not, it turns out, on the company’s very best one. That one is held back, reserved for a smaller and stranger set of problems nobody has solved before at all. I sat with that for a while. The tool built to survive a whole team’s whole week, in public, under the most sustained pressure any of their products face, wasn’t handed the sharpest blade in the drawer. It was handed the second-sharpest — which was apparently, entirely, enough. Which tells you something about where the two kinds of intelligence actually diverge: the merely-very-good model handles the desk staying clean for a week, in public, in front of a whole team, where one bad assumption made Monday and never revisited would be visible to everyone by Thursday. The truly new capability is being held in reserve for something else altogether.

I don’t have a tidy place to land this, and I’m suspicious of anyone who does. But here’s the closest I can get.

Imagine a three-Michelin-star chef — the kind of person who has spent thirty years learning to coax something transcendent out of a single scallop, who can tell you, by smell, that a stock has forty more minutes in it — standing at your stove on a Tuesday night making you a grilled cheese sandwich. It will, I promise you, be a very good grilled cheese sandwich. The bread will be evenly golden. The cheese will have reached some ideal, fully-considered state of melt. But almost none of what makes that chef extraordinary is actually being used to make it — none of the thirty years spent learning to hold forty things in mind at once without losing track of any of them, the exact skill, it occurs to me, that keeps a long, complicated project from quietly falling apart on day three. The technique is idling. The thirty years are in the room, present, available, and almost entirely beside the point, because a grilled cheese sandwich was never the place where thirty years shows up. It shows up somewhere else — in a dish you will never order, on a night you weren’t there.

What you got instead, on your ordinary Tuesday, was simply more than enough.

Categories
AI AI: Inference Semiconductors Uncategorized

5 Critical Management Lessons from the Founders at Etched

How two young founders are building what could become one of the most important companies in the AI era — and what their story teaches about leadership, execution, and building at the edge of the possible.

I recently listened to the latest Invest Like the Best podcast from Patrick O’Shaughnessey which was a remarkable conversation with Gavin and Rob, the founders of Etched, the company building specialized AI inference hardware that’s aiming to be radically better than existing solutions. Their story — starting as very young founders against massive skepticism, raising serious capital, and now shipping full rack-scale systems — is packed with hard-earned wisdom.

One of the comments Patrick makes at the beginning was how during his due diligence on the company he kept being told that semiconductor technology wasn’t a place for young people. You need seasoned, middle age experts to master this domain. Exactly not these founders.

Note: the following is based upon an AI’s analysis of the conversation transcript with me asking “What are the five most important management lessons from this conversation?” These lessons are relevant whether you’re leading a team, building a product, or simply trying to do meaningful work in our fast-moving world.

1. Velocity Compounds — Prioritize Speed Ruthlessly

In hardware, and increasingly in any deep-tech endeavor, speed isn’t just an advantage; it’s often the deciding factor.

Etched didn’t just design a chip — they built the full inference solution (chip, board, power delivery, interconnects, cold plates, and production processes) in parallel. They sent engineers to live in Bangalore for months to unblock vendors. They ran 24/7 shifts and did massive pre-work (including putting full chip designs on FPGA clusters) so that when the silicon finally arrived, they had working inference in racks in just 40 days.

Key takeaway: Look for every opportunity to parallelize. Accept higher short-term costs if they buy meaningful time. As they put it, “You win by shipping.” The best part is often no part — and the best vendor is no vendor, when vertical integration lets you move faster. Velocity, velocity, velocity.

2. Build Teams with Legends + High-Drive Talent

One of the most distinctive parts of their approach is how they recruit. They seek out “Legends” — people who have done the hardest versions of the problem before (like the engineer who built Nvidia’s HGX and DGX systems) — and pair them with exceptionally driven, somewhat naive high-performers who refuse to accept conventional limits.

They use “project-based recruiting,” mapping the hardest technical problems ever solved and persistently pursuing the actual people who did the real work. Their culture self-selects for people willing to move their families to San Jose to bet on two young founders taking on the world.

Key takeaway: For breakthrough work, average talent doesn’t suffice. The combination of deep experience and raw, first-principles energy creates magic. Invest heavily in finding and retaining these people — even if it takes 20 conversations. You can also learn a lot if the best in the world talent turns down the opportunity to work with you!

3. Assume It’s Possible, Then Solve the “Unsolvable” Problems

Repeatedly in their story, experts told them certain things were impossible. Their response? Assume it is possible and figure out how.

The most striking example was a clock domain crossing issue that required aligning signals to within 50 picoseconds — something many engineers said couldn’t be done. People quit. They solved it in about two weeks during a very dark period.

Key takeaway: When you hear “impossible,” treat it as the beginning of the investigation, not the end. Cultivate a “find a way” mindset across the team. The moments when things feel hopeless are often when the most important progress happens. I’m constantly struck by how often persistence results from simply realizing (or assuming) that something is actually possible.

4. Production Is the Real Product

Etched’s mantra is “Production is the product.” They obsess over not just technical performance but manufacturability, supply chain resilience, serviceability, and the ability to scale to gigawatts.

They made deliberate choices around process nodes and memory to avoid zero-sum competition. They built their own factory processes and test infrastructure early. Future designs are being simplified specifically for faster production cycles and higher reliability at massive scale.

Key takeaway: In any business that hopes to reach real scale, think end-to-end from the beginning. Technical excellence without production excellence is just a prototype. Optimize for output (tokens, units, whatever your metric is) at volume. There’s a lot of “zero to one” thinking here.

5. Bet Big and Stay Existentially Focused

Building in semiconductors requires enormous capital. Etched raised roughly $100 million early on when they were still very young and pre-tapeout — after most traditional investors had passed. They knew half-measures wouldn’t work.

This existential focus (this one product determines whether the company lives or dies) creates a different level of intensity that attracts talent, suppliers, and customers who believe.

Key takeaway: Match your ambition with appropriate resources and commitment. Clear existential stakes help filter for the right people and partners. In a world of distractions, singular focus on what truly matters is a superpower.

Final Thoughts

Gavin and Rob’s story is the combination of technical sophistication and deep human resilience. They faced a tough personal battle with cancer (in Rob’s case), widespread doubt, brutal technical challenges, and fundraising pressure — and kept moving forward with curiosity, determination, and humility.

In an age of AI and accelerating technology, the ability to build teams that can solve seemingly impossible problems at speed may be one of the most valuable capabilities a leader can develop. Their example reminds us that the future belongs not just to the smartest, but to those who can execute with urgency while maintaining clear principles. Velocity, velocity, velocity.

Categories
AI Learning Photography

Autopilot

“Superb photographs are not just taken with cameras. They come from within you, your eyes, your mind, your heart, not ice cold equipment.” Fan Ho

There’s a half-second on the street, somewhere between seeing a frame and shooting it, that used to take me whole minutes. Early on, with a camera in my hands on the streets of San Francisco or on the subway platforms in New York, I’d see something — light falling a certain way, a gesture about to resolve into a gesture — and I’d think my way through it. Assess the composition or the angle. Worry about the background. By the time I’d worked it out, the moment might be gone, replaced by some lesser version of itself.

That doesn’t happen to me anymore, and I couldn’t tell you when it stopped. Somewhere along the way the thinking disappeared and the shooting stayed. I see the frame and the shutter goes, and only afterward, looking at the file, do I understand what I saw. I didn’t explicitly decide to skip the thinking. It just stopped showing up, the way a habit eventually stops asking your permission. Or how driving a car becomes second nature.

I think about this because of a problem the AI labs have been calling continual learning. The AI models we use are like brilliant interns. They can solve a hard problem at nine in the morning and a harder one by five, and they’ll astonish you doing it. But every session starts over from zero. Whatever they got right on Tuesday evaporates by Wednesday, the way a dream is gone by the time you’ve found your slippers.

The industry’s first answer was to give them a longer memory — let the window hold the whole case file in front of them, all the time. This works for a while, the same way it would work for me on the street if I stopped and re-derived the exposure math for every frame. But that isn’t how I shoot anymore. I don’t have the math open. I have what’s left after thousands of frames did the math for me and then got out of the way.

Based on some exploration I did this morning using AI I found three different AI research efforts that are now chasing that gap, from different angles, none of them all the way there.

A team out of Stanford and NVIDIA built something called TTT-E2E, which lets a model keep adjusting its own internal weights while it reads — not just holding the page in front of it, but being changed by the page, a little, as it goes. It runs thirty-five times faster than the brute-force method of remembering everything, because it isn’t remembering everything.

Google’s research arm published something called Nested Learning around the same time, built on the idea that a mind isn’t one system learning at one speed, but several systems nested inside each other — some updating by the minute, some by the year.

And a scrappier strand of work called self-distillation has models teaching cheaper versions of themselves, not by handing over a transcript, but by training the cheaper model to arrive on its own at whatever the well-informed version would have concluded.

None of this is what happens when I make a photo. Not yet. But it’s aimed at the same gap I live in every time I shoot before I understand what I’m shooting. The gap between having the math and having the eye.

I once asked Doug, a good friend who’s spent as many days on the street as I have, how he knew when to press the shutter. He didn’t have an answer, not really — just a shrug, and something about the moment feeling complete before he could explain why. That shrug took him years to earn. He didn’t keep the years. He kept the shrug.

And then a few years ago Doug did something I still don’t fully understand. He abandoned digital and went back to film. Not for any project, not for the look of it — he could get that in post if he wanted it. He went back to the actual mechanics: loading a roll, metering by hand, often using a tripod, etc. I needled him about it some, the way you’d needle a cigarette smoker who’d taken up a pipe instead, as if the inconvenience were the point. He told me he wanted to slow down, and that film was the only thing that reliably made him do it. Twelve frames and then you stop and reload and you can’t fix it later. The very friction he’d spent decades shooting his way out of, he went looking for again, on purpose.

I don’t know what to do with that, except to notice that he’s the same man who can give me the shrug and also the man who walked back toward the thing the shrug had replaced. Maybe that’s the part the labs haven’t gotten to yet, underneath all the vocabulary of weight updates and meta-learned initializations. Compression is the whole point, until the day it isn’t.

Note: This line of thinking started with a recent essay by Dwarkesh Patel on what he calls continual learning. It’s become a real focus of his thinking about how we get to a better future with AI.

See: https://www.dwarkesh.com/p/the-next-paradigm

Categories
AI Anthropic Economics Stanford

Weak Links, Powerful Ideas

I’ve been thinking about bottlenecks. Not the frustrating kind you encounter in traffic or while debugging code, but the deeper structural constraints that determine how progress unfolds in our lives, organizations, and economies. A single slow step can limit an entire system, regardless of how rapidly everything else improves.

It’s an idea that feels especially relevant today. While AI capabilities continue to advance at a remarkable pace, real-world productivity gains often appear far more gradual.

Enter Chad Jones—the Stanford economist whose work has become increasingly important for anyone trying to understand AI’s long-term economic impact. This week he announced that he will join the Anthropic Institute on leave from Stanford beginning June 30.

The move is noteworthy not simply because of who Jones is, but because of the ideas he brings with him.

The Economist Who Sees Growth Through Tasks and Bottlenecks

Chad Jones (Charles I. Jones) is the STANCO 25 Professor of Economics at Stanford Graduate School of Business. He has been one of the leading scholars studying long-run economic growth: how ideas accumulate, why innovation matters, and why growth rates have remained relatively stable even as the number of researchers worldwide has expanded dramatically.

His influential work helped explain a central paradox of modern economics: adding more researchers does not automatically produce ever-faster growth because, over time, new ideas become increasingly difficult to discover.

More recently, Jones has turned his attention to artificial intelligence. Papers such as A.I. and Our Economic Future and his 2026 collaboration with Chris Tonetti, Past Automation and Future A.I.: How Weak Links Tame the Growth Explosion, examine how advances in automation may reshape economic growth in the decades ahead.

The central insight is deceptively simple:

Economic output is ultimately constrained by its weakest components.

Weak Links: The Economic Version of Amdahl’s Law

Anyone with a background in computing will recognize a familiar pattern.

Amdahl’s Law tells us that even if part of a program becomes infinitely fast, overall performance remains constrained by the portion that cannot be parallelized. Accelerating 90 percent of a workload by a factor of a million still leaves the remaining 10 percent as a hard limit on total speedup.

Jones’ “weak links” framework applies a similar logic to the broader economy.

In task-based models where tasks are complements rather than easy substitutes, every task matters. Extraordinary progress in a handful of areas does not automatically translate into extraordinary gains for the system as a whole if critical bottlenecks remain.

Historically, a large share of productivity growth has come from automation—the transfer of tasks from human labor to rapidly improving machines and capital. Jones and Tonetti argue that much of past productivity growth can be understood through this lens. The breakthrough is not merely building better machines; it is expanding the range of tasks that machines can perform.

The AI Timeline Paradox

Looking ahead, the same logic applies to AI.

Even as advanced models automate larger portions of cognitive and physical work, growth may continue to be constrained by:

  • Tasks that still require human judgment or participation
  • Regulatory and institutional frictions
  • Physical-world coordination challenges

As a result, Jones’ modeling suggests that economic growth may accelerate substantially while still unfolding more gradually than either enthusiasts or skeptics expect.

This perspective offers a useful middle ground between two popular extremes: the belief that transformative AI-driven abundance is imminent and the belief that AI’s impact will prove largely illusory. Progress can be both real and constrained. The chain remains only as strong as its weakest link.

From Theory to Practice

One reason Jones’ work resonates with me is that it extends beyond economics.

Many successful builders and leaders instinctively operate according to a weak-links philosophy. Whether in engineering, manufacturing, logistics, or organizational design, the greatest gains often come from identifying the single constraint that limits the system and focusing disproportionate effort on removing it.

Consider how Elon Musk has approached challenges at Tesla, SpaceX, and xAI. Across very different domains, a recurring pattern emerges: identify the binding constraint, concentrate resources there, remove it, and then move to the next bottleneck.

Jones’ framework provides an economic explanation for why this approach can be so effective. In systems composed of complementary tasks, relieving a key constraint can create benefits that ripple throughout the entire system.

Why This Resonates

What I find most compelling about Jones’ work is its intellectual balance.

It neither dismisses the remarkable capabilities emerging from frontier AI systems nor assumes that technological progress automatically translates into social or economic transformation. Instead, it directs attention toward the frictions, constraints, and complementarities that determine how change actually unfolds.

At a time when conversations about AI often oscillate between utopian abundance and existential catastrophe, this framework offers something rarer: a disciplined way of thinking about progress.

The weak-links perspective reminds us that the future may be shaped less by spectacular breakthroughs than by our ability to identify and address the constraints that prevent those breakthroughs from creating widespread value.

A Chain and a Compass

There is a quiet power in recognizing weak links—whether in economies, organizations, projects, or our own lives. The places where progress feels slow or frustrating are often where the greatest leverage resides.

Jones’ research provides a language for understanding those constraints, and his move to the Anthropic Institute suggests that some of the most important conversations about AI’s future may increasingly take place at the intersection of research, policy, and real-world deployment.

For that reason alone, this is a development worth watching.

If you’re interested in exploring the underlying ideas, I recommend starting with Jones’ recent papers on his Stanford faculty site, along with Anthropic’s announcement of the Institute and its mission.

Categories
AI

What the Lessor Keeps

Two airlines can fly the same airplane. Not airplanes of the same type — the same airplane, serial number and all, handed back at the end of a lease and reassigned, sometimes within weeks, to a competitor on another continent. AerCap owns more commercial aircraft than any airline on earth, and it leases them to airlines that spend their advertising budgets convincing passengers that flying them is a distinctive experience. The 737 MAX that wears Ryanair’s livery this year might wear Lion Air’s the next, repainted, recertified, its avionics untouched, its airframe indifferent to the change of ownership. The lessor does not care who is flying its asset. It cares that the asset comes back in airworthy condition and that the lease payments clear.

What the airline owns, in the sense that matters, is never the aircraft. It is the route network built up over decades of slot negotiations at constrained airports. It is the maintenance log — every inspection, every part swapped, every anomaly a mechanic in Singapore flagged in 2019 that turned out to predict a fatigue crack nobody else had seen yet. None of that travels with the airplane when the lease ends. It stays behind, compounding, in systems the airline built and the lessor never touches.

Karl Mehta, who has spent a career inside enterprise software watching this kind of asymmetry repeat itself, put a version of it plainly: a model is a brain you rent, and you and your competitor rent the same one. The formulation has the compression of something that has been tested in a few dozen meetings before it found that sentence. It is also, structurally, the airplane story. Anthropic and OpenAI and Google are AerCap. They retain residual value on enormous capital assets — clusters of GPUs depreciating on a schedule, weights trained at a cost that only a handful of balance sheets in the world can absorb — and they lease access to those assets by the token, to anyone who can pay, including, in the same afternoon, two companies trying to put each other out of business. The model does not know whose prompt it is answering. It has no loyalty file. It has, in fact, no memory at all, in the ordinary sense of the word — each call begins exactly where the last one ended for everybody, which is nowhere.

The asymmetry that airlines exploit is the one available here too, and it sits one layer up from the engine. Call it the embedding store, the vector database, the fine-tuning corpus, the retrieval index — the terminology varies by vendor, but the function is constant. It is the accumulated, indexed residue of every customer interaction a company has had, structured so that the rented brain can be handed the relevant fragment of it at the moment of each new call. A bank’s fraud model and a competing bank’s fraud model can call the identical foundation model, route through the identical API, and arrive at entirely different verdicts on the identical transaction, because one of them is retrieving against eleven years of labeled chargebacks specific to its own card portfolio and the other is retrieving against four. The intelligence rented by the hour is, for practical purposes, a commodity, priced down toward marginal cost the way jet fuel is priced — everyone pays close to the same number per unit. The memory is not a commodity. It cannot be, because it is not for sale; it is the institutional record of what has already happened to you, and no amount of capital lets a competitor buy a copy of your chargeback history any more than it lets them buy your maintenance logs.

This produces a particular kind of corporate vertigo, which Mehta’s sentence is really addressing. For three or four years the industry conversation about artificial intelligence has been a conversation about models — which lab’s was larger, which benchmark moved, which release cycle a company should anchor its roadmap to. That conversation rewards being an early and aggressive lessee. But a lessee relationship, however aggressive, does not compound into anything a competitor cannot eventually also lease. The compounding, when it happens, happens in the layer below the API call: in how cleanly a company has structured the record of its own customers, its own failures, its own edge cases, so that the rented brain, plugged in fresh every morning with no memory of yesterday, can be handed exactly the right fragment of yesterday and made to look, for a few hundred milliseconds, like it has been there all along.

A hospital chart has two kinds of entries. There is the vital-signs strip clipped to the bed rail — temperature, pulse, blood pressure, checked every four hours and replaced every four hours, because a reading from yesterday tells the night nurse nothing about the patient in front of her right now. And there is the permanent record in the file downstairs: the allergy that nearly killed him in 2019, the surgery, the medication history going back a decade, written once and never overwritten, because that record is exactly as valuable ten years from now as it is today. Nobody confuses the two charts. Nobody staples last Tuesday’s blood pressure into the permanent file. The hospital figured out, long before anyone digitized it, that memory is not one problem. It is two, and they fail in opposite directions if you run them through the same system.

Most teams building the layer Mehta is describing make exactly that mistake — they staple everything to the same chart. The shorthand for it is dumping everything into a vector database and praying, and it is worth asking why that particular error is so popular. The answer is that it feels like progress: embeddings go in, something resembling memory comes out, and the team moves on to the next sprint without confronting the harder question, which is what kind of memory it just built.

Short-term memory is the vital-signs strip — everything the model needs to finish the task in front of it and nothing it needs after. A customer-service exchange in progress, the order number already mentioned, the fact that this is the second call today, belongs here. So does the scratchpad of a multi-step agent: the search results just pulled, the file just opened, the partial answer being assembled before it commits. The test is not how important the information is but how long it stays true. A customer’s mood this minute is real and gone in twenty minutes; storing it permanently is like stapling yesterday’s temperature reading into the permanent file, undated, until the chart tells you nothing about fever and everything about clutter. Short-term memory should live in the context window itself, or a session-scoped cache, and it should be allowed to die when the session ends. The sin is not forgetting it. The sin is remembering it forever.

Long-term memory is the file downstairs, and it does not come in one shape any more than that file does. The first shape is semantic memory — facts. A customer’s account tier. The chargeback history that decides, in fractions of a second, whether this morning’s transaction clears. Facts belong in a database with a schema, not a vector store, because a fact has a right answer and a vector store gives you an approximate neighbor. Ask a vector index what tier a customer is on and it hands you the five most semantically similar sentences in the corpus — one correct, four merely correct-sounding. Ask a schema the same question and it tells you, because that is what the schema is for.

The more sophisticated shops are already building the seam between the two, rather than picking one and living with its blind spot. A knowledge graph keeps the relationships a schema is good at — this customer, that account, this chargeback, in fixed and queryable connection to one another — while still letting a retrieval layer search across it by meaning rather than by exact key. The approach has a name now, GraphRAG, and the name matters less than what it concedes: that facts and resemblance are different operations, and the honest fix is to run both and let each one answer the kind of question it’s actually suited for, not to force a single index to pretend it can do both jobs at once.

The second shape is episodic memory — what actually happened. The specific conversation last March in which the customer explained, at length, why the previous fix didn’t work. The exact sequence of an agent’s failed attempt at a task, preserved so the next attempt doesn’t repeat it. This is where the vector store finally earns its keep, because an episode isn’t an exact-match lookup, it’s a resemblance — has anything like this come up before — and a vector index, built to find the nearest thing to a fuzzy question, is the right tool for that question and almost no other. The error was never using a vector store. The error is using only a vector store, for facts as well as episodes, on the theory that one hammer with sufficient cosine similarity can stand in for the whole toolbox.

The third shape is the rarest, and the one teams forget to build at all: procedural memory, which is not a fact and not an episode but a skill — the model’s learned sense of how this company writes a refund email, escalates a complaint, formats an invoice. Style is the visible half of it. The other half is harder to see and matters more: the rails the model is forced to run on before it ever gets to choose a word. A refund above some threshold routes to a human, no exceptions, because the workflow says so, not because the model was persuaded to think so on this particular call. An agent that touches a production database does it through a reviewed function with a fixed set of permitted calls, not through whatever query it improvises in the moment. None of that lives in a prompt, and none of it lives in the model’s weights either. It lives in code — the orchestration layer, the permissioning, the state machine the agent is required to pass through — and it is procedural in the oldest sense of the word: not a memory of what to say but a memory of what is and isn’t allowed to happen, enforced whether or not the model that day feels like remembering it. It doesn’t live in a database at all. It lives in fine-tuning, in carefully maintained house-style examples, and in the surrounding scaffolding of guardrails and permitted actions, and it changes slower than the other two, the way a surgeon’s hands carry both technique and caution years after the specific patients are forgotten. A company that has built rich semantic and episodic memory but skipped this layer has a model that knows everything about its customers, writes in exactly the right voice, and is one well-crafted prompt away from doing something the company never agreed to.

The real argument here is not which database serves which layer — that part is plumbing, and plumbing changes every eighteen months. The argument is that memory has to be triaged the way the hospital triages it, with something deciding on purpose what survives the session and what doesn’t, rather than writing every token of every interaction into the same undifferentiated store and trusting retrieval to sort it out later. A vector database with no triage in front of it is not a memory system. It is a landfill with a search function, and it will retrieve the wrong eleven-month-old conversation with the same confidence it retrieves the right one, because nobody wrote the part of the system whose only job is deciding what belongs on which chart.

The lessor’s airplane, repainted, will fly for someone else next year. The route network will not. Neither will the schema that knows a customer’s tier on contact, nor the index that remembers the conversation from last March, nor the fine-tuned hand that knows, without being told twice, how this company writes a refund email. These are the things that do not come back at the end of the lease, because they were never on it.

Categories
AI AI: Large Language Models AI: Transformers Authors Podcasts Writing

The Billboard

The fog was still sitting on the hills when I put in my earbuds and headed out.

Sebastian Mallaby was talking about billboards.

Tim Ferriss had asked him the question he asks everyone: if you could put anything up there, for millions of people to see, what would it be? Mallaby has spent years inside the minds of the people who shaped modern finance — the hedge fund managers, the venture capitalists, the builders of things that changed how the world moves money. He has more material than most people accumulate in a lifetime. He could have said anything.

He said: Prepare your mind.

I kept walking. The houses were quiet in the particular way they get when school lets out for summer — no buses, no car doors, no kids at the corner. Somebody’s sprinklers were running.

The phrase comes originally from Louis Pasteur, who understood something that most people don’t: that chance is not democratic. It does not distribute itself evenly among those who wait. It finds the people who are ready. Chance favors the prepared mind. Pasteur said it, and then he proved it, and then the rest of us spent a century and a half learning it was true.

What struck me about Mallaby’s answer wasn’t the phrase itself. It was the way he said it had kept appearing in his research, surfacing in different decades and different worlds, like a message the material kept trying to send him.

He told the story of Arthur Patterson at Accel Capital. Before a new technology arrived, Accel would work through the implications — what company needs to be built, what founder fits the moment, what the right pitch looks like. So when an entrepreneur finally walked in, when the situation was live and competitive, they already knew ninety percent of what they were hearing. They could move fast because they had already moved slow.

That’s preparation as institutional practice. But Mallaby found the phrase again in a different register entirely, embedded in a single human moment that has always seemed to me like one of the hinge points of our era.

He was interviewing Ilya Sutskever, asking him why he had seen it so quickly.

In 2017, a paper called Attention Is All You Need appeared online. It described a new architecture for neural networks — the transformer — that would eventually rewrite the terms of what artificial intelligence could do. On the day the paper went up, Sutskever read it. And then he ran. He went down the corridor to find his collaborator Alex Radford and told him to stop what he was doing. Everything. Stop. We are going to build a language model on this architecture.

Not someday. Now.

Mallaby asked him how he had seen it so clearly, so fast. And Sutskever’s answer, in its essence, was the same two words: prepared mind.

He had been thinking about the problem of modeling sequential data since his PhD in Canada. For years he had been carrying a question the field hadn’t answered yet. And when the answer appeared — when the transformer showed up on a website one ordinary day — he didn’t have to reason his way toward it. He recognized it. The solution arrived and found a mind that had been waiting for it, that had already cleared space for it, that was already arranged around the shape of exactly this kind of answer.

This is what preparation actually is. Not the accumulation of facts. Not readiness in the generic sense, the vague self-improvement sense. It is the long, patient cultivation of a specific question, held close and kept alive until the answer has somewhere to land.

Mallaby chose that phrase for his billboard because it kept finding him — in the venture capital world, in the AI world, across decades and disciplines and very different kinds of genius. The prepared mind is not a personality trait. It is a practice. It is the work you do before the work arrives.

The sprinklers had clicked off by the time I turned back toward home. The fog was starting to lift off the hills. I was thinking about what I had been preparing for, whether I even knew.