Geoffrey Hinton says a big language model runs on about 1% of your brain's connections and still ends up knowing more than you: "So in your brain, you have a hundred trillion connections, roughly speaking. Okay. That's a lot. And you only live for about two billion seconds. That's not much." "If you compare how many seconds you live for, with how many connections you've got, you have a whole lot more connections than experiences." "Now with these neural nets, it's sort of the other way round. They only have of the order of a trillion connections. So like 1% of your connections, even in a big language model, many of them fewer, but they get thousands of times more experience than you." "So the big language models are solving the problem with not many connections, only a trillion. How do I make use of a huge amount of experience?" "And back propagation is really, really good at packing huge amounts of knowledge into not many connections." "But that's not the problem we're solving. We've got huge numbers of connections, not much experience. We need to sort of extract the most we can from each experience." Two to three billion seconds is the whole budget. Everything you know, you learned inside it. So evolution built you to squeeze a lot out of very little. Hinton's point is that a language model has the opposite problem and the opposite fix, and backprop turned out to be extremely good at that fix. Worth noticing what this predicts about failure. A system running on 1% of your wiring and thousands of times your experience is not going to fail the way you do. You fail from having seen too few examples. It fails from compressing too many into too little, and the compression is where the errors get made. That is a strange thing to be deploying into hospitals and courts with no way to inspect it. We test these systems by asking them questions, which tells you what came out. Nobody can yet look at a trillion connections and say what got packed in. - Geoffrey Hinton, Nobel laureate and Turing Award winner, on StarTalk (@StarTalkRadio) with Neil deGrasse Tyson.

the results from this week's @a16z build survey are in! here's a glimpse... your dream angel investors: - @eladgil - @pmarca - @elonmusk - @chamath - @saranormous - @danielgross people you think will start the next big co: - @PhilipLakin - Dir of AI Transformation at Zapier, prev Co-founder/CEO of NoCodeOps - @kalinowski007 - Hardware & Physical AI leader; prev OpenAI, Meta, Oculus VR, Apple, Stanford - Rob Anderson - Chief of Staff at Atomic Machines; prev SpaceX, Miso Robotics, Caltech - Vaishali Parekh - PM at Meta, prev Pinterest - Jeffrey Wang - Member of Technical Staff at OpenAI; prev Harvard Teaching Fellow companies you'd join if you quit your job tomorrow (skipping the classics): - @anduriltech - @probookai - @fal - @DecagonAI - @blueorigin the rest of the results are hidden - only shared with people who filled out the survey link in thread to get a head start on next week's survey!

This week in SF with award-winning filmmaker Darren Aronofsky @DarrenAronofsky, @openai researcher Noam Brown @polynoamial, and Dylan Golden president of Primordial Soup We discussed the future of filmmaking, building creative models, and Hollywood and Silicon Valley’s long partnership & recent tension Put on the by the one & only @eladgil, & @emmzaoui

Inference-time model routing based on legal practice area dramatically improves agent performance. So do other types of "blended intelligence": - collaborative model teams, - advisor-executor patterns, - and model routing based on inferred user preference. While these methods appear promising, they come with substantial risks and eval challenges. We're experimenting with all of them at Harvey. Read more from our Head of Legal Research @ItsJulioPereyra:

There’s an insane amount of alpha in: - People in SF leaving SF to see how people and companies around the world are using AI (hint: most are still using it primarily as search) - People outside SF spending even two weeks in SF to see how people and companies are operating here (hint: many are already living in the future) This may be obvious to some, but the gap right now is staggering. The global diffusion of AI is a massive opportunity.

For 9+ years we’ve been building simulators, data engines, data pipelines, scenario editors, RL environments, what's now world models... and dozens of other products. Now we’ve connected them all under one agentic layer. It is not just another harness but also the broader system to safely design, develop, and deploy Physical AI. I continue to believe a modern, AI-first development environment is one of the biggest bottlenecks to putting intelligence on a billion machines. Unsurprisingly, we’ve been using it internally for a while now and the impact has been profound in almost every measurable way. As you might know, our old office was on the corner of Pioneer and Dana – and calling it Pioneer would be too cliche. Introducing Dana :)

We are hiring for @Harvey’s model training team. This team will help Harvey expand from the application layer into the model layer and from legal into high end knowledge work more broadly. We are hiring AI researchers of all seniority, particularly those with experience post-training frontier or open source models. Our program is centered around large-scale model training, synthetic data generation, long horizon reinforcement learning, and rigorous evaluation in real world deployments. We are scaling-pilled and believe that nothing beats the combination of larger models and better training data. We’ve been able to generate incredibly realistic legal environments and validated that this allows us to post-train open source models to achieve frontier performance with agents. We plan to scale up these data generation and training efforts significantly across legal to start, and eventually other verticals. As a researcher, you will have access to thousands of GPUs and unique training data from our product and customer relationships. Your research will inform Harvey’s product strategy and power AI used for some of the most economically and societally impactful work in the world.

How to keep AI spend flat while token usage grows exponentially: Not with friction and spend alerts. With better defaults, routing, and caching. Better Defaults (not Usage Caps) – Engineers can choose any model they want, but defaults matter. We’re experimenting with defaulting to open weight models like GLM 5.2 and Kimi 2.7 through our LLM gateway, while still encouraging engineers to choose the right model for the task. 91% of our employees were never hitting their usage caps, so instead of lowering caps and driving up alerts, we're moving to cheaper defaults. Note that code reviews use a diversity of models, so they can check each other's work. Better Routing – In our custom harnesses, we preprocess prompts and route to the best model for the job, considering cache hits and model pricing. For instance, you may want a frontier model for planning, but not for execution where they can be overkill. Ultimately, humans shouldn't be choosing models - AI can automate this task. Better Caching – Cache misses are the easiest way to drive your cost up. All of our requests are cache aware, so we’re reusing a warm cache wherever possible. For example, our cache hit rate went from 5% → 60% in LibreChat once properly implemented. Keep Context Lean – Start fresh sessions when switching tasks. Scope file context narrowly. Disconnect unused tools. Don't just compact. The goal isn't fewer tokens used, it's fewer tokens wasted. Better Visibility – Our engineers can use as many tokens as they want, from whatever model they want, but we’ve made usage visible – and the more you spend on AI, the more impact we expect. The goal isn't to suppress usage. It's to build the infrastructure that makes exponential growth sustainable. Putting this into practice has cut our AI spend nearly in half, while our token usage continues to grow.