Tim Middleton-Sally 🇺🇸

Distilled recap of the back-and-forth with Jensen on export controls: Dwarkesh: Wouldn’t selling Nvidia chips to China enable them to train models like Claude Mythos with cyber offensive capabilities that would be threats to American companies and national security? Jensen: First of all, Mythos was trained on fairly mundane capacity and a fairly mundane amount of it by an extraordinary company. The amount of capacity and the type of compute it was trained on is abundantly available in China. Dwarkesh: With that, could they eventually train a model like Mythos? Yes. But the question is, because we have more FLOPs, American labs are able to get to this level of capabilities first. Furthermore, even if they trained a model like this, the ability to deploy it at scale matters. If you had a cyber hacker, it's much more dangerous if they have a million of them versus a thousand of them. Jensen: Your premise is just wrong. The fact of the matter is their AI development is going just fine. The best AI researchers in the world, because they are limited in compute, also come up with extremely smart algorithms. DeepSeek is not an inconsequential advance. The day that DeepSeek comes out on Huawei first, that is a horrible outcome for our nation. Dwarkesh: Currently, you can have a model like DeepSeek that can run on any accelerator if it's open source. Why would that stop being the case in the future? Jensen: Suppose it optimizes for Huawei. Suppose it optimizes for their architecture. It would put others at a disadvantage. As AI diffuses out into the rest of the world, their standards and their tech stack will become superior to ours because their models are open. Dwarkesh: Tesla sold extremely good electric vehicles to China for a long time. iPhones are sold in China. They didn't cause some lock-in. China will still make their version of EVs, and they're dominating, or smartphones, they're dominating. Jensen: We are not a car. The fact that I can buy this car brand one day and use another car brand another day is easy. Computing is not like that. There's a reason why x86 still exists. There's a reason why Arm is so sticky. These ecosystems are hard to replace. Dwarkesh: It's just hard to imagine that there's a long-term lock-in to the Chinese ecosystem, even if they have this slightly better open-source model for a while. American labs port across accelerators constantly. Anthropic's models are run on GPUs, they're run on Trainium, they're run on TPUs. There are so many things you can do, from distilling to a model that's well fit for your chips. Jensen: China is the largest contributor to open source software in the world. China's the largest contributor to open models in the world. Today it's built on the American tech stack, Nvidia’s. Fact. All five layers of the tech stack for AI are important. The United States ought to go win all five of them. in a few years time, I'm making you the prediction that when we want American technology to be diffused around the world—out to India, out to the Middle East, out to Africa, out to Southeast Asia—on that day, I will tell you exactly about today's conversation, about how your policy ... caused the United States to concede the second largest market in the world for no good reason at all.

the universal rule of working with ex-government/intelligence community/confidential jobs people is that the more sketchy they are about what they did, the greater the odds they were an unimportant cog real actors have better cover stories

As a former Special Forces Warrant Officer with multiple rotations running counterinsurgency ops—both hunting insurgents and trying to separate them from sympathetic populations—I’ve seen organized resistance up close. From Anbar to Helmand, the pattern is familiar: spotters, cutouts, dead drops (or modern equivalents), disciplined comms, role specialization, and a willingness to absorb casualties while bleeding the stronger force slowly. What’s unfolding in Minneapolis right now isn’t “protest.” It’s low-level insurgency infrastructure, built by people who’ve clearly studied the playbook. Signal groups at 1,000-member cap per zone. Dedicated roles: mobile chasers, plate checkers logging vehicle data into shared databases, 24/7 dispatch nodes vectoring assets, SALUTE-style reporting (Size, Activity, Location, Unit, Time, Equipment) on suspected federal vehicles. Daily chat rotations and timed deletions to frustrate forensic recovery. Vetting processes for new joiners. Mutual aid from sympathetic locals (teachers providing cover, possible PD tip-offs on license plate lookups). Home-base coordination points. Rapid escalation from observation to physical obstruction—or worse. This isn’t spontaneous outrage. This is C2 (command and control) with redundancy, OPSEC hygiene, and task organization that would make a SF team sergeant nod in recognition. Replace “ICE agents” with “occupying coalition forces” and the structure maps almost 1:1 to early-stage urban cells we hunted in the mid-2000s. The most sobering part? It’s domestic. Funded, trained (somewhere), and directed by people who live in the same country they’re trying to paralyze law enforcement in. When your own citizens build and operate this level of parallel intelligence and rapid-response network against federal officers—complete with doxxing, vehicle pursuits, and harassment that’s already turned lethal—you’re no longer dealing with civil disobedience. You’re facing a distributed resistance that’s learned the lessons of successful insurgencies: stay below the kinetic threshold most of the time, force over-reaction when possible, maintain popular support through narrative, and never present a single center of gravity. I spent years training partner forces to dismantle exactly this kind of apparatus. Now pieces of it are standing up in American cities, enabled by elements of local government and civil society. That should keep every thinking American awake at night. Not because I want escalation. But because history shows these things don’t de-escalate on their own once the infrastructure exists and the cadre believe they’re winning the information war. We either recognize what we’re actually looking at—or we pretend it’s still just “activism” until the structures harden and spread. Your call, America. But from where I sit, this isn’t January 2026 politics anymore. It’s phase one of something we’ve spent decades trying to keep off our own soil.

Honestly, Ben Affleck actually knowing AI and the landscape caught me off guard, but as a writer, makes sense. Great takes across the board.

Right now all design problems for LLM-based apps feel like a choice between: - Solve it for 100% of the problem space at X% reliability - Solve it for Y% of the problem space at 100% reliability (And you don’t know the values of X and Y up front)

Sharing an interesting recent conversation on AI's impact on the economy. AI has been compared to various historical precedents: electricity, industrial revolution, etc., I think the strongest analogy is that of AI as a new computing paradigm (Software 2.0) because both are fundamentally about the automation of digital information processing. If you were to forecast the impact of computing on the job market in ~1980s, the most predictive feature of a task/job you'd look at is to what extent the algorithm of it is fixed, i.e. are you just mechanically transforming information according to rote, easy to specify rules (e.g. typing, bookkeeping, human calculators, etc.)? Back then, this was the class of programs that the computing capability of that era allowed us to write (by hand, manually). With AI now, we are able to write new programs that we could never hope to write by hand before. We do it by specifying objectives (e.g. classification accuracy, reward functions), and we search the program space via gradient descent to find neural networks that work well against that objective. This is my Software 2.0 blog post from a while ago. In this new programming paradigm then, the new most predictive feature to look at is verifiability. If a task/job is verifiable, then it is optimizable directly or via reinforcement learning, and a neural net can be trained to work extremely well. It's about to what extent an AI can "practice" something. The environment has to be resettable (you can start a new attempt), efficient (a lot attempts can be made), and rewardable (there is some automated process to reward any specific attempt that was made). The more a task/job is verifiable, the more amenable it is to automation in the new programming paradigm. If it is not verifiable, it has to fall out from neural net magic of generalization fingers crossed, or via weaker means like imitation. This is what's driving the "jagged" frontier of progress in LLMs. Tasks that are verifiable progress rapidly, including possibly beyond the ability of top experts (e.g. math, code, amount of time spent watching videos, anything that looks like puzzles with correct answers), while many others lag by comparison (creative, strategic, tasks that combine real-world knowledge, state, context and common sense). Software 1.0 easily automates what you can specify. Software 2.0 easily automates what you can verify.