remember when Andrej Karpathy released autoresearch, where an agent runs experiments in a loop on…
Machina
Visit on X@EXM7777160K followers
18,751 views
remember when Andrej Karpathy released autoresearch, where an agent runs experiments in a loop on its own?
well i packed his method into one prompt that connects Claude, ChatGPT, and Grok into a shared loop inside Roma
- writes a check that can actually fail, first
- splits the task into parts across your agents
- routes research to Grok and architecture to Claude
- loops each bot until it passes the check
- syncs results to one shared to-do list
- turns every mistake into a rule for next time
- calls you only for the step that can’t be undone
throw it at Roma with the task you’ve been putting off for a week, and you’ll see the difference in one run
it’s insane how well this thing works, let me know if you want the full setup
Comments
Comments from the original post on X (Twitter)
No comments yet
