Derek

The next step for autoresearch is that it has to be asynchronously massively collaborative for agents (think: SETI@home style). The goal is not to emulate a single PhD student, it's to emulate a research community of them. Current code synchronously grows a single thread of commits in a particular research direction. But the original repo is more of a seed, from which could sprout commits contributed by agents on all kinds of different research directions or for different compute platforms. Git(Hub) is *almost* but not really suited for this. It has a softly built in assumption of one "master" branch, which temporarily forks off into PRs just to merge back a bit later. I tried to prototype something super lightweight that could have a flavor of this, e.g. just a Discussion, written by my agent as a summary of its overnight run: github.com/karpathy/autor… Alternatively, a PR has the benefit of exact commits: github.com/karpathy/autor… but you'd never want to actually merge it... You'd just want to "adopt" and accumulate branches of commits. But even in this lightweight way, you could ask your agent to first read the Discussions/PRs using GitHub CLI for inspiration, and after its research is done, contribute a little "paper" of findings back. I'm not actually exactly sure what this should look like, but it's a big idea that is more general than just the autoresearch repo specifically. Agents can in principle easily juggle and collaborate on thousands of commits across arbitrary branch structures. Existing abstractions will accumulate stress as intelligence, attention and tenacity cease to be bottlenecks.

This weekend we hosted two challenging hackathons on two different continents! Excited to see so many people passionate about protein design. We’re looking forward to experimentally test all of those new proteins in our wet lab next, stay tuned for the results 🇩🇪 Berlin Bio × AI Hackathon The Protein Design Track we organized had 6 teams competing by designing binders against 15-PGDH, a gerozyme whose rising activity with age drives muscle loss, neurodegeneration, and joint decline. We saw teams bringing their own models to the task, improving them in 24 hours, while others used out-of-the-box solutions. Some people even created new dashboards with Claude Code integrated for direct feature requests. And others focused on critical challenges for building a biotech: delivering the designer drugs, market analysis, exploring other binder modalities. We’re now testing the best 150 proteins in our lab and will release the experimental results next month! 🇺🇸 SF bioArena Hackathon - Humans vs AI agents A single-day hackathon where human teams and AI agents competed side by side in designing proteins. The target was TREM2, a receptor involved in Alzheimer’s disease. 9 human teams competed against 6 agents, including Anthropic’s Claude Sonnet 4.6 and xAI’s Grok 4.1. 100 designs were selected for experimental validation in our wet-lab. We’ll reveal who won, agents or humans, soon! 🧬Want to participate in one of our competitions? Join the RBX1 challenge on @proteinbase - the submission deadline is at the end of this month. Thanks a lot to @la_Payette_ , Stefan Hristov, @burgshrimps, and the rest for organizing the Berlin Bio hackathon and to @katyenko for the bioArena one in SF.

In Silico to In Vitro is our most ambitious hackathon yet. We’re teaming up with @adaptyvbio to run a challenge where scientists generate novel binders in silico, with top designs built and tested in the lab. Join us if you're in SF on 2/28: luma.com/a6t92ohv Co-sponsored by @RowanSci, @openrouter and @modal, with generous compute support.

@Yuchenj_UW You guys do realise that humans can also post on this site right?