Why OpenAI is merging Codex and ChatGPT and the future of knowledge work | Andrew Ambrosino

When building working software gets cheap, choosing what to keep becomes the scarce skill. That is the argument of Andrew Ambrosino, who leads product and engineering for the Codex desktop app at OpenAI, in a Lenny's Podcast interview.

Ambrosino says product work used to assume implementation was expensive, so teams wrote documents, research and prototypes first to de-risk what engineers built. In Ambrosino's account, models now let anyone stand up a feature quickly, so many people at OpenAI build their own versions of the same idea. The costly step becomes curation: what to fold together, how to frame it, whether it belongs in another feature. Ambrosino calls that taste.

Ambrosino does not call documents obsolete. A prototype that looks production-ready can wrongly signal that research and business questions are settled, so Ambrosino says to pick the medium deliberately: a document for a vague area, a prototype to stress-test an interaction. This is one practitioner's account of a single team.