OpenAI built a Codex factory to keep ChatGPT fast

OpenAI has described the Performance Factory, a system it built to keep ChatGPT fast. It is a set of Codex agents running in the cloud that hunt for places where a change has made ChatGPT slower, investigate them and help fix them.
Speed has been a running theme in OpenAI’s engineering posts. Earlier this year the company said reads in Habitat, the storage behind ChatGPT, fell from roughly 120 ms to roughly 10 ms.
It is also the latest example of OpenAI running its own house on Codex. The company previously said it built Pando, a new Git backend for its main code repository, in 10 weeks with Codex.
Up next
Anthropic hits Nasdaq in the biggest IPO ever, Video
Dario Amodei rang the Nasdaq bell for Anthropic, ticker ANT. CBS News says the IPO raised about $81 billion at a $1.5 trillion value, beating SpaceX’s $75 billion. Would you buy ANT shares on day one or wait?
Related stories
- OpenAI’s Pando cut monorepo push latency nearly 5x
- OpenAI’s lawyers and recruiters now work through Codex
- Codex runs OpenAI workflows through Runme
- OpenAI’s Habitat cut read latency from 120 ms to 10 ms
- OpenAI: agents do the work, people still have the ideas
- What the Codex sandbox blocks and what it can’t
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

