1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵

5/ Autoresearch helped us develop Flint. Seeing Flint contribute useful research itself is an encouraging early step. We're starting to explore how far that loop can go, and how to develop it safely. Meet Flint: xitter.cf/cong_ml/status/2112949… Try it locally: xitter.cf/cong_ml/status/2112966…

4/ Flint finished close to historical Haiku 4.5, with Sonnet 4.6 still ahead at the same experiment budget. Our 40B model uses just 2.5B parameters per token. We didn't expect it to get this far on a research task we'd associated with powerful proprietary models.

3/ Here, Flint mainly tuned hyperparameters for a very specific training budget. It explored batch size, depth, learning rates and schedules—testing combinations and building on what worked. The progress came from adapting the recipe to the constraints.

2/ Can Flint improve language-model training with the same compute? Five minutes on an H100 per NanoChat experiment: change code, train, measure, try again. NanoChat was our bread and butter while developing our Eureka Machine. Nice to be back: recursive.com/updates/first-…

1/ Can a model that runs on a laptop also drive GPU training research? Recursive Flint, our new 40B open model, improved NanoChat across 700+ experiments—finishing close to our historical Claude Haiku 4.5 results. We were pleasantly surprised. Here's what it tried 🧵