honored to be on the TIME 100 health list!

Introducing Simile. Simulating human behavior is one of the most consequential and technically difficult problems of our time. We raised $100M from Index, Hanabi, A* BCV, Andrej Karpathy Fei-Fei Li Adam D'Angelo Guillermo Rauch scott belsky among others.

x.com/i/article/2022…

Very excited about the "First Proof" challenge. I believe novel frontier research is perhaps the most important way to evaluate capabilities of the next generation of AI models. We have run our internal model with limited human supervision on the ten proposed problems. The

The golden age of discovery is here garryslist.org/posts/demis-ha…

SOLVE EVERYTHING... youtube.com/watch?v=EbCZ4Q…

We went from AI systems that struggled to do grade school math to AI systems that can solve research-level math problems in just a few years. I agree with Jakub this is perhaps the most important eval now. I am also pretty sure the main reaction will be "it's not that hard" :)

highest returns to intelligence tokens

imo, people still haven't fully realized how to leverage claude skills as composable primitives, especially when fully managed/hosted developers should feel like kids in a candy store right now

Predicting the future accurately is the best measure of intelligence

Our team recently published an article in Harvard Business Review on why context (not model access!) is the true competitive advantage in the age of AI. When everyone has access to the same AI models, the same tools, and the same vendor ecosystem, context becomes the

FDA formalizes its shift to requiring only one pivotal trial (via NEJM) and analysts are already pointing at companies this could help -- another pro-pharma push for Makary, and another 180 for Prasad who wrote extensively in his academic life against trials focusing on anything

x.com/i/article/2024…

The most underrated AI opportunity for startups: automating how companies talk to each other. Within a company, AI is making rapid progress. But across companies? The world still runs on emails, phone calls, PDFs, and FTP. A thread 🧵

Computer use models shouldn't learn from screenshots. We built a new foundation model that learns from video like humans do. FDM-1 can construct a gear in Blender, find software bugs, and even drive a real car through San Francisco using arrow keys.

Galen has assembled one of the most mythical teams including Neel Redkar and yudhister, who have exceeded what frontier labs spent billions of dollars on with a fraction of the resources. Classic case of a small group of brilliant kids running circles around corporate labs

what is the right prior for computer use? the SI team made a big bet that it is video + action conditioning, not text + images the result is FDM-1, and its 2 hour 30fps video context mogs everything else out there

.OpenAI is building an army of FDEs for Life Sciences and leaving the rest to consultants/SIs specifically because the upside of abundant intelligence isn't evenly distributed. Applying abundant intelligence to the latent complexity of life sciences has the ability to produce

Ray Kurzweil doubles down: By 2029, computers will know everything humans know—and we'll explode scientific progress like never before. He predicts AGI (human-level AI) arrives in 2029 (a timeline he's held since 1999, now widely seen as credible amid rapid AI leaps) and the

A fantastic article that reiterates what I’ve been saying for a year: AI only drives meaningful ROI for a business when it’s custom built for the way your employees already do their work People always ask me how Varick competes with “Agentic AI SaaS” and I always respond with