Ex-OpenAI researcher Jerry Tworek launches Core Automation to build the most automated AI lab in the world

Tworek led the team behind OpenAI's Codex, which makes his departure particularly pointed given that OpenAI just expanded Codex into a flagship agentic product. He isn't joining an established competitor; he's attempting to build a lab whose core thesis is that AI research itself should be automated, which is a structural bet against the current human-researcher-in-the-loop model that every major lab still runs on.

This is the third notable OpenAI research or product leader departure in roughly a week, following Bill Peebles (Sora) and Kevin Weil, both covered here in mid-April. Taken together, the exits suggest OpenAI's consolidation around Codex and a narrower product surface is pushing out people whose work didn't survive the prioritization cut. Core Automation's pitch also sits in tension with the MIT Technology Review argument from April 16 that competitive advantage in AI is shifting toward operational infrastructure rather than raw research capability. Tworek is betting the opposite: that whoever automates the research loop first wins the capability race outright.

Watch whether Core Automation publishes any technical writing or recruits additional OpenAI alumni within the next six months. If it does neither, this reads as a holding announcement rather than an active research program.