In a rare podcast interview, Greg Brockman opened up about the 72 hours after Sam Altman was fired from OpenAI, detailing the chaos that nearly destroyed the company. Brockman described receiving the board call, quitting the same day, and designing a backup entity called Phoenix at Altman's house the next morning. He also credited Ilya Sutskever's tweet with shifting the situation.

Brockman traced the technical roadmap back to OpenAI's original Napa offsite, which produced a three-step plan the company followed for a decade. That offsite also revealed why OpenAI had to abandon its pure nonprofit structure, pushing it toward a for-profit model that enabled the growth trajectory ChatGPT now follows.

Looking at OpenAI's current state, Brockman said it is hard to know what percent of the company's code is not written by AI. OpenAI also stopped showing reasoning traces in ChatGPT because they create false impressions of reliability. The conversation framed a compute-constrained world as the central variable determining who accesses AGI.