What's this about?
Sequoia interviews OpenAI's Core Automation leads Jerry Tworek and Rohan Anil on automating AI research itself — agents running experiments, analyzing results, and proposing next steps.
What exactly is an "automated AI lab"?
The idea that the research loop — hypothesize, run, analyze, iterate — can itself be delegated to AI systems, with humans setting direction. It's the mechanism behind claims of accelerating AI progress, and the context for OpenAI's math results this week.
Why does it matter?
Tworek is one of the more senior researchers to describe this concretely; worth a listen for how much of the loop is *already* automated (Jul 29 episode).