AI agents can now do research on their own
Technology
Andrej Karpathy, one of the early members of OpenAI, just dropped an open-source tool called autoresearch that lets AI agents do LLM research on their own: no human babysitting needed.
With just three files and one GPU, these agents tweak code, run five-minute experiments, check their results, and keep improving, all while you sleep.
How agentic AI could change science
This is a glimpse into the future of AI that works at "human speed" or faster.
Karpathy calls it a simple demonstration, but it's already got people talking about how agentic AI could change how science gets done.
If you're curious about where tech is headed, or just want to see what happens when research runs itself, this is one to watch.