Alexander Embiricos: evaluation

Transcript context

…So what I'm hearing is the initial version of Codex was almost too far in the future. It's like a remote in the cloud agent that's coding for you asynchronously. And what you did is, "Okay, let's actually come back a little bit, let's integrate into the way engineers already integrate into IDs and locally and help them on ramp to this new world," Totally. And it was quite interesting because we dogfood product a ton at OpenAI. So dogfood as in we use our own product. And so Codex has been accelerating OpenAI over the course of the entire year, and the cloud product was a massive accelerant to the company as well. It just turns out that this was one of those places where the signal we got from dogfooding is a little bit different from the signal you get from the general market because at OpenAI, we train reasoning models all day and so we're very used to this kind of prompting and think upfront, run things massively in parallel and it would take some time and then come back to it later asynchronously. And so now when we build, we still get a ton of signal from dogfooding internally, but we're also very cognizant of the different ways that different audiences use the product. That's really funny. It's like live in the future, but maybe not too far in the future. And I could see how everyone at OpenAI is living very far in the future, and sometimes that won't work for everyone.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.