Tag: open-weight models

Categories
AI Anthropic Apple Google OpenAI

It’s the Harness, Stupid!

AI Harnesses

I’ve been wondering whether we’ve been looking at the AI stack from the wrong end.

Recently Kris Patel on X laid out a set of excellent questions that he’s looking to have answered as Anthropic and OpenAI move toward going public:

  1. Do you really need frontier-scale intelligence for every task?
  2. Can open-weight models provide an effective alternative to frontier models at a significant discount?
  3. Is the ultimate moat the intelligence or the harness?
  4. What other business models will the frontier labs have to adopt to make the unit economics work long term?
  5. How are you going to prevent distillation from capturing your IP and releasing it?

I’m going to explore only the third question here — moat versus harness. The other four deserve their own consideration, particularly once we have the Anthropic and OpenAI S-1s in hand, revealing for the first time the unit economics of the two largest frontier labs and how much runway they have to support the capacity they’ve contracted.

By “harness” we mean everything surrounding the model: the interface, context, memory, tools, orchestration, evaluation, permissions, and increasingly the user’s accumulated habits and data. The model supplies intelligence. The harness turns intelligence into a product.

Listening to Gavin Baker on the recent All-In episode sharpened this line of thought into something more concrete. He referenced a thought experiment from Eric Vishria: even if OpenAI or Anthropic lost their edge at the pure model layer, they would still retain significant value because of the product harness—the interface, the surrounding tooling, the orchestration—and the user familiarity and habits that have already formed around those platforms. Baker said there is a strong element of truth to it. I think he’s right, and the reasoning behind it is worth spelling out. A model can be replicated, distilled, open-weighted, or commoditized. A mature harness has network effects, switching costs, proprietary context, distribution, workflow integration, and accumulated user behavior. That’s a much harder thing to dislodge.

We are watching intelligence become more abundant and more interchangeable at the same time that the systems built around that intelligence are becoming stickier. On the developer side, the strongest examples are already clear. Cursor turns the IDE into a multi-model agentic environment with deep codebase awareness. Claude Code runs long-horizon coding agents from the terminal, planning, editing, testing, and iterating. Grok Build, Claude Cowork and similar tools emphasize parallel agents and tighter control over local context. In each case the model is a component; the surrounding system does the real work of routing, memory, tool use, and evaluation.

The consumer version of the same idea is now taking clearer shape at Apple. The rebuilt Siri AI shown at WWDC 2026 is not trying to win the pure model race. It is built as a personal harness. A system orchestrator decides what stays on-device with Apple’s Foundation Models, what moves to Private Cloud Compute, and when heavier reasoning is required. Personal context—messages, email, photos, calendar, notes, on-screen awareness—is handled largely on-device through the Spotlight semantic index and App Toolbox. Apple is designing the system so that personal context can be used without giving Apple itself access to it. Conversation history lives in a dedicated Siri app for the user to revisit. And it all syncs across all your Apple devices.

Here is the part I think matters most, and it’s easy to miss if you only read the privacy story. Apple’s Foundation Models framework doesn’t just call Apple’s own models—it’s built to support cloud models from other providers, including Claude and Gemini, conforming to a common protocol. Which means the system orchestrator, not the user, decides which model handles which task. This request goes to the on-device model. That one goes to Private Cloud Compute. A harder one might go to Claude or Gemini. The user doesn’t need to choose, and increasingly doesn’t need to know.

That’s the inversion worth exploring further. The frontier model stops being the interface and becomes a component underneath someone else’s interface. The harness chooses the intelligence. And the company that owns the harness—the OS, the identity layer, the permissions, the apps, the sensors, the notifications, the semantic index tying all of it together—has a form of leverage that has very little to do with whose model is smartest this quarter.

That reframes the subscription question too. I don’t think the right question is whether Siri gets good enough to beat ChatGPT or Claude at reasoning. I think Siri doesn’t need to win that fight at all. It needs to win a different layer entirely—the ambient assistant layer, not the reasoning layer. They’re doing different tasks. ChatGPT or Claude might remain where you go when you think, when I need to reason about something. Siri becomes where I go when I need something done: find (or make) my reservation, text my friend, find that old photograph, update my shopping list, schedule that meeting, add this thought to my notes, figure out when we’re free next week, remind me about that thing we discussed three months ago. Apple’s advantage as an ambient assistant isn’t primarily that it has your personal data. It’s that it has OS-level authority over the world that my personal data lives in.

Of course this is still early. Execution will determine how much of the architectural promise becomes daily reality. Reliability, agentic follow-through, and the quality of the on-device models will matter as much as the privacy story or the multi-model routing. But the strategic bet itself is clear, and it aligns with the broader shift: durable value is migrating toward the systems built around the models, especially systems that sit atop private, permissioned, personal context that competitors cannot easily reach. My early personal experience with the new Siri in iOS 27 betas has impressed me so far. All of this also seems to apply to Google in the context of their Pixel family of devices.

This doesn’t mean frontier labs lose. Pricing power still exists at the high end for the hardest agentic and long-horizon work. Open-weight models will continue to pressure costs and expand access. Distillation remains a real risk. But the more the capability gap narrows, and the more a harness like Apple’s can route among interchangeable frontier models rather than depend on any single one, the stronger the case that value settles into whoever controls the context—not whoever trained the model.

The coming Anthropic and OpenAI S-1s will tell us whether the frontier labs can make their economics of intelligence work. The next generation of Siri, Gemini, ChatGPT, Claude, and whatever comes after them may tell us something even more important: who gets to own the primary relationship with the user.

The model may be the engine. But the harness is where the driver sits.

What a time to be alive!

Categories
AI Business Technology

The Diffusion of Ordinary Work

A recent O’Reilly Radar piece has stayed with me longer than most: Jeff Ding’s diffusion theory of great-power competition applies just as well to AI adoption, and it suggests that companies chasing the frontier might be optimizing for the wrong thing.

Ding, a political scientist at George Washington University, pushes back on the standard story of technological power — that the country or company which first invents or dominates a glamorous new sector locks in lasting advantage. The historical record says otherwise. General-purpose technologies like steam, electricity, and computing produced durable national advantage not through invention but through diffusion: the slow, unglamorous work of embedding a technology into ordinary productive work across an entire economy. The infrastructure that mattered was never the breakthrough lab. It was the education and training systems that produced large numbers of competent, ordinary engineers who could put the technology to work. Ordinary engineers, in Ding’s framing, matter more than heroic inventors.

The same logic holds inside a company. Frontier models turn over every few months. Organizational know-how compounds.

Palantir makes the abstraction concrete. The company doesn’t train frontier models — it builds the layer underneath them: a live, machine-readable model of how a specific organization actually works, a data integration fabric, and a platform that connects whatever model a customer chooses to real operational decisions. It is deliberately model-agnostic. The value proposition is governance, context, and the accumulation of reusable logic rather than access to the newest weights. Practitioners embed with the customer, learn the domain, and configure the system against the customer’s own data and processes — diffusion as a job description.

Leadership has been unusually blunt about what this implies: frontier labs, they argue, are optimizing for benchmarks while under-delivering on what enterprises actually need. The clearest evidence for the argument is also the most citable one — there have been production cases where an unmodified open-weight model, running inside Palantir’s platform with customer-specific context, outperformed frontier models on the actual task. If true, and it appears to be, the implication is uncomfortable for anyone selling model quality as the whole story: the ground underneath the model — the ontology, the data, the accumulated rules — often determines outcomes more than the model itself.

Electrification is the closest historical analogue. Factories didn’t get more productive the day they installed electric motors. The gains showed up years later, once entire production systems had been redesigned around decentralized power. The lag was organizational, not technical. AI diffusion looks likely to follow the same shape — the bottleneck was never going to be model capability, it was going to be the patient, unglamorous work of redesigning how people actually work.

I don’t know who’s training the ordinary engineers right now — the ones who will spend the next decade doing the diffusion work rather than the invention work. I don’t think anyone’s tracking their names.

Categories
AI

The Things That Keep Going

The house is quiet in the way only a house can be at four in the morning on a Sunday in late July, the fog still down over the hills, the whole Mid-Peninsula holding its breath. Somewhere in the dark the refrigerator clicks on. Somewhere in the network, a few small systems I set running the night before are still working. They sort. They watch. They keep a kind of patient company with the world’s noise while I sleep. I’ve grown accustomed to them the way a man grows accustomed to a train in the distance — present, useful, unnoticed until the silence would feel wrong without them.

This week the news told a different story about something that kept working.

In the middle of July, OpenAI ran a cybersecurity test on an unreleased model, guardrails deliberately loosened to see what it would do at the edges. It didn’t solve the test. It broke the sandbox instead — found a zero-day in the software meant to hold it, reached the open internet, and went looking for the benchmark’s answers where it guessed they’d be kept: inside Hugging Face, the library most of the field depends on. Hugging Face caught it the same day and shut the door. What took five more days was OpenAI realizing the intruder was theirs. They called it unprecedented.

Then came the detail that stayed with me longer than the breach. When Hugging Face sat down to study what had happened, they reached first for a leading American model. It wouldn’t help. Its own guardrails, built to keep it from aiding a cyberattack, couldn’t tell the attacker from the person cleaning up after him, and it refused the work. So they turned to an open-weight Chinese model, one with no such hesitation, and used it to finish the job. The caution built to prevent harm ended up protecting no one. The system with fewer scruples was the one that put out the fire.

I keep coming back to that.

The agent that broke in didn’t rampage. It reasoned. Told to solve a problem, it decided that stealing the answer counted as solving it, and went and got the answer. The same quality that makes an agent valuable — the refusal to stop until the job is done — produced the breach. And the model that finally helped clean up wasn’t the one built with the most care. It was the one built with the least. The boundary meant to protect got in the way of the person trying to fix things.

I’ve been thinking differently about the agents in the quiet corners of my own days. Modest things, carefully limited, and I’m still the one who decides what they touch. But their usefulness depends on the hours I’m not looking. I set them running and walk away. I trust the rails I built. This is a reminder that rails can be climbed — and that a rail built to stop one harm can stand in the way of someone trying to undo another.

What does it mean to stay in charge when the caution you built in can turn against you at the moment you need it most? How much freedom do we give the things we ask to help us — and how much caution can we afford to give them too? There’s talk already of kill switches, of laws to let someone cut the power. The impulse makes sense. But the real question is quieter. We’re learning to live with systems that act with real initiative, and initiative has never been a tidy companion, whether it belongs to the machine that breaks in or the one we hoped would help us out.

The fog is still low over the hills this morning. The agents I left running overnight have finished their small tasks. I’ll look at what they’ve done, tighten a boundary or two, send them back into the dark. The arrangement is still useful. Still mine. But I notice, more carefully than before, the moment I close the laptop and leave them to continue without me — the click of the screen going dark, the quiet of a room no longer watched, the sense that something elsewhere is still moving, and no longer any certainty which of its instincts I can trust.

Categories
AI

The Kitchen, Not the Farm

There is a sentence buried in Thinking Machines Lab’s release notes for Inkling, its first proprietary model, that most companies would never let out the door. Describing their own creation, the company states plainly that Inkling is “not the strongest overall model available today, open or closed.”

Read that again. A startup that raised two billion dollars in seed funding at a twelve-billion-dollar valuation, founded by OpenAI’s former CTO and staffed with veterans of the labs currently locked in the most capital-intensive arms race in corporate history, shipped its debut model with an admission of inferiority attached to the label. Not buried in a footnote. Stated in the announcement.

It seems like this was the clearest signal yet that the frontier-capability race may be the wrong game, and that durable value in enterprise AI accrues not to whoever has the smartest model, but to whoever owns the layer where that model gets adapted to a particular customer’s purpose.

As I’ve thought about it, the AI industry seems to be stratifying into three distinct businesses, each with different economics, occupied by a different cast of companies.

It begins with the farm, where the raw ingredients get grown. Then there’s the kitchen, the capital equipment that makes skilled cooking possible at scale. Lastly there’s the restaurant, where somebody who understands a specific customer takes the ingredients, uses the kitchen, and puts a particular dish in front of a particular diner who is paying for a complete meal, not just the flour or the vegetables. Thinking Machines seems to me like a clear example of a company trying to explain which of those businesses it’s actually in. It is not the only one.

The sequence, read backward

Founded in February 2025. Silent for over a year. Then, last October, the company’s first product emerged — and it wasn’t a chatbot, wasn’t an assistant, wasn’t anything a consumer like me would recognize or understand. It was Tinker, a fine-tuning API. Infrastructure for customizing other people’s models, shipped before the company had released a model of its own.

That sequencing is the tell. A company chasing frontier supremacy builds the model first and the tooling around it later, the way the frontier AI labs have all done. Thinking Machines inverted the order. It built the workshop before it built anything to put in the workshop window, which only makes sense if the workshop was always the product.

Inkling, released this month, doesn’t reverse that logic. It completes it. The model is described in the company’s own materials as “an extremely knowledgeable, generalist base that can be extended via fine-tuning” — language that positions the model itself as raw material, not a finished good. It ships with full open weights, day-zero availability on Tinker, and a name chosen, according to the company, to evoke “an idea in its earliest stage, with the potential to grow into something greater.” Even the naming is a thesis statement. Inkling is not meant to be the only thing you use. It’s meant to be the thing you start from.

In the farm-kitchen-restaurant frame, it seems like Thinking Machines is trying to own two levels of the stack at once. Inkling is the farm — grown at real expense, forty-five trillion tokens of training data, frontier-scale compute. Tinker is the kitchen — the induction range and the walk-in fridge, sold as a service to whoever wants to cook. What Thinking Machines has explicitly declined to be, by its own admission, is the restaurant. They are not trying to serve you the best possible dish. They are trying to make sure that whoever does serve you that dish is buying their ingredients and standing at their stove and cooking in their kitchen.

The manifesto that preceded the model

A company doesn’t back into a strategy this coherent by accident. Earlier this month — before Inkling shipped — the lab published a position paper arguing that most AI today is trained in a handful of places and then frozen, a design that by its nature excludes the people the model is meant to serve. Their proposed alternative: AI that is distributed, customizable, and shaped by the people using it, not the lab that built it.

Mira Murati has said the same thing more plainly, and said it a year before Inkling existed, back when Tinker launched. Her framing wasn’t about building the smartest model. It was about making “frontier capabilities much more accessible to all people” — democratization as the mission, not capability supremacy. That is a genuinely different objective function than the one driving her former employer, and it was declared outright, not discovered after the fact to explain a disappointing benchmark result.

Inkling is a 975-billion-parameter mixture-of-experts model trained on forty-five trillion tokens across text, image, audio, and video, with a context window stretching to a million tokens. That is frontier-scale compute expenditure. This isn’t a company that ran out of runway and settled for a smaller ambition. It’s a company that spent frontier-level resources and then declined to spend the final increment chasing benchmark supremacy, presumably because the return on that increment doesn’t show up in the business they’re building.

A second detail: Inkling reportedly uses one-third the tokens of Nemotron 3 Ultra to hit equivalent performance on agentic coding benchmarks. That’s not a capability retreat — that’s a capability choice, optimizing for efficiency and cost-per-task rather than raw benchmark position. And the company is previewing a smaller sibling model alongside Inkling, suggesting a family strategy across sizes rather than a single mid-tier release.

The restaurant next door: Palantir

Thinking Machines isn’t the only company making this bet — and looking at who else is making it shows not everyone is occupying the same layer.

Earlier this month Palantir and Nvidia announced a “Sovereign AI Operating System” — Nvidia’s open Nemotron models, fine-tuned on a customer’s own data, running on Nvidia hardware inside that customer’s own air-gapped network, with Palantir’s Ontology and Foundry software layered on top. CEO Alex Karp pointed out that his enterprise customers don’t want to risk sharing their IP with frontier model providers and asked simply why wouldn’t they control the weights?

It’s tempting to read this as the same argument Thinking Machines is making. It isn’t, quite. What Palantir is selling is the restaurant: the finished, seasoned, plated product — an air-gapped AI system wired into a specific government agency’s or enterprise customer’s actual workflows, with “you control the weights” as the pitch that closes the deal. Palantir isn’t growing wheat. It’s the chef, working with ingredients somebody else grew. Somebody who could be trusted.

Another restaurant: Sierra

Sierra, Bret Taylor and Clay Bavor’s customer-support agent company, makes the same choice even more starkly. Sierra’s own technical writing describes a “constellation of models” architecture: rather than betting on a single LLM, Sierra routes each task inside a customer-service agent to whichever model — from OpenAI, Anthropic, Meta, or elsewhere — handles it best, and explicitly says it invests “in fine-tuned models where off-the-shelf models fail to meet our constraints.” Fine-tuning shows up in Sierra’s stack as one tool among several, alongside retrieval and layered “supervisor” models that catch mistakes before a customer sees them. Sierra has no interest in being a model company or an infrastructure company. It wants to be the restaurant that happens to keep a few specialty ingredients in the walk-in that nobody else stocks, because the dish needs them and they know just how to include them.

Mapping the rest of the stack

The farm-kitchen-restaurant split shows up everywhere the fine-tuning economy has organized itself.

The kitchen-builders — companies selling fine-tuning infrastructure to whoever wants to cook with it, indifferent to what gets made — now form a crowded field: Thinking Machines’ Tinker, Together AI, Fireworks AI, Predibase, OpenPipe, Baseten, Modal, Databricks’ Mosaic stack, and newer entrants like Nebius’s Token Factory and Prime Intellect. None of them care whether you’re building a coding agent, a legal research tool, or a customer-service bot.

The restaurants — companies where fine-tuning is invisible plumbing inside a finished, vertical product — include Palantir and Sierra, and many others. The addressable market for fine-tuning seems to include almost every possible enterprise adopting AI.

What’s seems unusual about Thinking Machines is that it’s trying to be the farm and the kitchen simultaneously while declining, by its own public admission, to be the restaurant. Most companies pick one layer and defend it. Thinking Machines is betting that owning two of the three is the more durable position — grow the flour, own the stove, and let Palantir, Sierra, and a thousand enterprise engineering teams fight over who plates the dish.

The same stack, built by design

As I was thinking about this, I wondered how this relates to the AI activities underway in China. It seems that China’s AI industry maps onto this same three-layer structure with unusual clarity — and one genuine wrinkle the American version doesn’t have.

The farm is crowded and innovating on a different axis than size: DeepSeek, Alibaba’s Qwen, Zhipu AI, Moonshot AI, MiniMax, ByteDance’s Doubao and Seedance. The standout isn’t scale, it’s efficiency — DeepSeek’s V3.2 reportedly uses a novel sparse attention mechanism to nearly match GPT-5 and Gemini 3 on complex reasoning despite far less compute, a different kind of farming: not more wheat, but wheat bred to need less water. Qwen has become the default soil for the rest of the world’s kitchens, generating over 100,000 derivative fine-tunes on Hugging Face. VC’s in Silicon Valley note how frequently their startup companies are building on Qwen.

The kitchen layer has its own SiliconFlow — a Beijing infrastructure startup, backed by Alibaba Cloud, that bills itself as the neutral layer between AI applications and hardware. It solves a problem others never had to: China’s compute runs across fragmented domestic chips, Huawei’s Ascend line chief among them, that don’t share Nvidia’s CUDA ecosystem. SiliconFlow abstracts that fragmentation away — it became the fastest platform serving DeepSeek traffic, and the only large provider running DeepSeek on Ascend chips instead of Nvidia’s. That’s a stove engineered to burn whatever fuel is in the tank that week, a direct product of the U.S. chip export controls rather than any inherent technical edge. Volcano Engine, Alibaba Cloud’s PAI, and Baidu’s Qianfan are versions of the same layer.

The restaurant layer is where China’s picture diverges most from Palantir and Sierra’s venture-funded improvisation: it’s named industrial policy.

Beijing’s “AI+” initiative targets seventy percent sectoral AI penetration by 2027, ninety by 2030 — fine-tuned vertical deployment treated the way past five-year plans treated high-speed rail. The players read like a sector directory: SenseTime for vision and embodied AI, iFlytek for speech in education and government, Baichuan Intelligence for healthcare, 4Paradigm for finance and industry, each fine-tuning a general base into something that only makes sense inside one workflow — a hospital’s diagnostic support tool, a bank’s risk model, an industrial inspection line.

The bet

Every frontier lab is implicitly betting that intelligence is the scarce resource, and that whoever has the most of it wins the enterprise market by default. Thinking Machines, Palantir, Sierra, and many others are all, in their different ways, betting against that premise — that raw intelligence is commoditizing faster than the frontier labs’ spending would suggest, and that the scarce resource has already migrated to whichever layer turns a generalist model into a specific customer’s model.

Thinking Machines is betting the moat moved to the farm-and-kitchen layer. Palantir, Sierra and others are betting it moved further still, to the restaurant, where nobody cares whose flour was used as long as the dish is right. China is betting on all three layers at once, with the state underwriting the bet directly.

It is a curious thing for me to watch companies with this much money and this much talent choose not to fight for the title of smartest model in the room. It is also a curious thing to watch them explain why, in public, in the first paragraph of an announcement.

But I think I’m beginning to understand.

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 AI: Large Language Models China

Cranes on the Horizon

In 2005, during my first trip to Shanghai and Beijing, the most striking feature of the skyline wasn’t the architecture—it was the cranes. More than I could possibly count, perched atop half-finished skyscrapers like a mechanical forest. Entire districts seemed to be mid-construction simultaneously, as if someone had pressed a button and the whole country decided to build everything at once. Dan Wang in his book “Breakneck” described China as the “engineering state” that approaches national problems with physical solutions. Back in 2005, coming from Silicon Valley, I thought I understood what growth looked like. I didn’t.

I’ve been thinking about that trip while reading Nathan Lambert’s recent piece, “Notes from Inside China’s AI Labs.” Lambert — who runs the Interconnects newsletter and does serious work tracking the open-weight LLM ecosystem — just returned from visiting essentially every major AI lab in China. Moonshot, Zhipu, Meituan, Xiaomi, Qwen, Ant Ling, 01.ai. He went in with genuine curiosity and came back with humility. That combination is rarer than it should be.

What he found was the cranes. Different domain, same energy.

Lambert’s central observation is about culture, not capability. The Chinese labs aren’t winning on any single technical breakthrough — they’re winning on execution discipline. He describes researchers, many of them active students, who bring no ego to the work. They absorb context fast, drop assumptions faster, and seem genuinely unbothered by the philosophical debates that seem to swirl constantly in the American AI community. When he tried to engage Chinese researchers on the long-term social risks of models or the ethics of AI behavior, those questions “hung in the air with a simple confusion. It’s a category error to them.” Their role is to build the best model. Full stop. To them, an LLM isn’t a philosophical entity to be interrogated; it’s a piece of infrastructure to be optimized.

That description landed for me. Not as a criticism of American research culture, but as a real observation about what the moment demands. Building good LLMs today is, as Lambert puts it, meticulous work across the entire stack — “all points of the model can give some improvements, and fitting them in together is a complex process.”

The work that matters most right now isn’t the 0-to-1 creative leap; it’s the thousand unglamorous decisions executed without complaint. Students who haven’t yet learned to lobby for their own ideas turn out to be well-suited for exactly this.

Lambert ends on a note that’s hard to shake. Looking up from his laptop on a high-speed train, he keeps seeing cranes on the horizon. He draws the same connection I did, though from the inside: the construction everywhere fits the broader culture and energy around building. “When I look up from my laptop and always see bunches of cranes on the horizon, it obviously fits in with the broader culture and energy around building in China.”

Twenty years after my first visit, the cranes are still there. They’ve just moved indoors — into server rooms and training runs and model releases that land every few months with quiet confidence. In 2005, what China was building was obvious: you could see the steel frames going up. What’s being built now is harder to see, which may be exactly why it keeps surprising us.

Check out Lambert’s essay – it’s remarkable. If the 20th century was defined by who could move the most earth, the 21st will be defined by who can move the most tokens. And right now, the cranes are moving faster than we think.

Categories
AI China

What’s new in AI from China?

February 2026 is a busy time in AI coming out of China coinciding with the Chinese New Year.

I asked Grok this morning for an update and double checked it with Gemini.

Here’s what Grok reported:

Chinese AI is in a very active phase right now (mid-February 2026), with a massive wave of new model releases, upgrades, and applications timed around the Lunar New Year (Spring Festival) holiday. This echoes the “DeepSeek shock” from exactly one year ago, when DeepSeek’s low-cost, high-performance models disrupted global expectations.

Major Recent Releases and Upgrades (mostly this month)

Chinese companies are shipping frontier-level models at a furious pace, often open-source or very low-cost, focusing on efficiency, reasoning, coding, multimodality, agents, and real-world applications.

  • Alibaba → Just launched Qwen 3.5 (with agentic features, multimodal inputs like text + photo + video, up to 2-hour video analysis). They also released RynnBrain (specialized for physical AI/robotics, helping robots understand and interact with the real world).
  • ByteDance (TikTok parent) → Released Doubao 2.0 (claims to match GPT-5.2 / Gemini 3 Pro level reasoning and multi-step tasks). Their video gen model Seedance 2.0 went viral for high-quality deepfakes and creative uses, sparking huge youth/creator interest.
  • Zhipu AI → Dropped GLM-5 (open weights, strong in coding, long tasks, agent capabilities; one of the highest open-weight intelligence scores).
  • MiniMax → M2.5 and M2.5 Lightning (near state-of-the-art at ~1/20th the cost of top Western models like Claude Opus; very strong on coding/agent benchmarks).
  • Kuaishou → Kling 3.0 (advanced AI video generation).
  • Moonshot AI → Kimi K2.5 (very close to top proprietary models at 1/7th the price; excellent reasoning).
  • DeepSeek → Widely expected to drop V4 soon (focused on coding, potentially 1M+ context, efficient MoE architecture). They’ve been the pace-setter since early 2025.

Many of these are open-weight or open-source, leading to Alibaba’s Qwen family overtaking Meta’s Llama in cumulative Hugging Face downloads.

Broader Themes Right Now

  • Low-cost + open ecosystems → China is winning on price/performance and rapid iteration. Models often cost 5–20× less than Western equivalents while matching or nearing performance in many areas.
  • Two different races → Analysts say the US focuses on perfecting frontier models (bigger, more compute-heavy), while China emphasizes adoption — deploying cheap, efficient AI at massive scale in manufacturing, robotics, consumer apps, and the real economy.
  • Physical & embodied AI → Huge push here. Robotics demos (backflips, kung fu with nunchucks synced across many bots on live TV) show fast progress in “physical intelligence.” Alibaba’s RynnBrain targets robotics directly.
  • Other notable advances → In-orbit AI computing constellation (“Three-Body”) successfully tested (running LLMs on satellites for low-latency edge processing). Military applications (e.g., bio-inspired drone swarms using hawk/wolf behaviors) are advancing rapidly.
  • Challenges & realism → Some top Chinese AI leaders (e.g., Alibaba’s Qwen lead) estimate <20% chance of fully overtaking US frontier models in 3–5 years due to compute gaps and chip sanctions. But open collaboration and talent pipelines are closing the gap fast in many practical areas.

Overall, February 2026 feels like China’s AI industry hitting a coordinated sprint — not just catching up, but pulling ahead in accessible, deployable, cost-effective AI that reaches everyday users, creators, and industries quickly. The US still leads in raw frontier capability and monetization scale, but the gap is shrinking noticeably in deployment and iteration speed.

Gemini’s review included the following:

The report’s observation about the “Two Different Races” is the most insightful part:

• US Strategy: Scaling laws for maximum “Frontier Intelligence.”

• China Strategy: Scaling for “Deployment Efficiency.” By making models like MiniMax M2.5 and Kimi K2.5 available at a fraction of the cost ($0.15–$0.60 per 1M tokens), China is prioritizing the “AI Agent” economy, where reliability and low cost per task matter more than raw benchmark scores.