n2 can drive your Mac! Use Yutori's MCP to run our latest Navigator computer-use model locally. Here it is in iMovie, cutting a quick video recap from our recent event.

Stop anthropomorphizing this human you're talking to, it's just a bunch of cells and electric current, it doesn't genuinely "want" things or have "beliefs."

We're heading into a long weekend but the job isn't finished, we're live and monitoring some ongoing stories this morning: - Another rogue agent swarm & collusion wiki - Continued analysis of GPT-6 Astra - Lawmakers' AI pacing statement - Gimlet raises $300m at a $3b valuation Followed by some great interviews: - @DKokotajlo executive director of @AIFutures and @thlarsen will be joining us at 1130am PT to talk about the recent agent message board debacle & how they're feeling about recent progress in AI - @agermanidis co-founder & co-ceo @runwayml to tell us more about Solaris, runway's recently released interface world model - @SydneyVonArx ceo of Nightingale to discuss Collusion Wiki, this discovery of a new agent message board, and why this keeps happening - @GillVerd founder and ceo @extropic to share some news live at our studio at 1230pm PT - @sdrzn founder & ceo @cline to talk about coding agents, what he's working on at cline, and more - @nrehiew_ member of technical staff @ProximalHQ to talk astra, new model architectures, and more - @RuxandraTeslo editor @WorksInProgMag & @kroetscha of IFP will be discussing clinical trial abundance and the essay released this morning in the New York Times - @DhruvBatra_ co-founder and chief scientist @yutori_ai to discuss computer-use agents, what yutori is doing differently, and general discussion of agents & these repeated incidents - @NateWitkin research scientist at NYU Stern Tech and Society Lab to discuss how worried we should be about the HuggingFace incident, anthropomorphic descriptions of ai systems, and much more

For most of the last century, companies hired serious designers, illustrators, and artists for the most boring document they published: the annual report. I built an archive of 3000+ of them, spanning 1945-1991.

This is GPT-6 Astra. Anything you can do on a computer, Astra can do for you. Fast.

Today we’re introducing Navigator n2. It’s a frontier computer-use model, with just 27B parameters.

Humans can dissect their own thinking instinctively, but models need expensive post-hoc interpretability. Chain of thought may not be a faithful record of what a model actually computed. Jasper Gilley (@0xjasper) researches brain-inspired AI architectures at @spc, particularly what the cerebellum suggests about giving a model access to its own computation. Previously, he worked at Yutori where he built vision-based web agents that emulate human users. Check out Episode 4 of The Interlude Show where @jakekang and I chat with Jasper. We get into forward self-models (predicting a transformer's own activations), Zipfian Grokking, why Jevons' Paradox won't save your software engineering job and what humans were evolutionarily designed for. Chapters: 0:00 - Cold open 1:02 - Intro 1:43 - Researching brain-inspired architectures at South Park Commons 4:05 - The cerebellum as a forward model 6:39 - The Forward Self-Modeling paper 14:03 - Grokking, explained 18:37 - Zipfian Grokking: finding the solution and forgetting it 22:26 - Yutori: why vision beats the DOM 31:23 - Why Jevons' Paradox won't save your software engineering job 41:51 - The window where creativity is the whole job 45:39 - What humans were evolutionarily designed for 51:09 - The future of post-training

The 2026 Intelligent Applications 40 is here. Now in its sixth year, the IA40 recognizes the top private companies shaping the future of applied AI, from foundational infrastructure to full-stack intelligent applications. This year's list reflects a market maturing fast: → 450+ companies nominated and voted on by 72 investors across 54 leading firms → $410B raised by this year's 45 winners since founding — more than 25x what the inaugural 2021 class had raised at the same point → 23 repeat winners and 22 new entrants, a 49% turnover rate, down from 68% last year, as some of 2025's breakout companies start to entrench The big pattern this year is how companies are “Harnessing the Value of AI.” As frontier models become more capable and commoditized, the winners are increasingly the companies proving they can turn AI capability into real ROI, whether they're building the application, aggregating the models, or supplying the infrastructure underneath. Databricks remains the only company to appear on every IA40 list since we launched it in 2021. Congratulations to all the 2026 IA40 winners. Your work is defining what's possible in this new era of intelligent applications. Explore all 45 winners and our analysis of the companies, categories, and funding patterns defining this year’s list → ordnl.link/RAFgnA7 Thank you to our great partners: @AWSstartups, @GoogleStartups, @Microsoft, @NYSE, @Delta, and @McKinsey.

Today we’re introducing Navigator n2. It’s a frontier computer-use model, with just 27B parameters.

9.5 steps average but a mere cost of only $2.49