OpenAI's Codex lead praises Ant's new coding agent for matching Codex background computer use
soumitrashukla9 · x · 2026-09-09
Thibault Sottiaux, who leads OpenAI's Codex, says Ant's new Claude Code-style coding agent ships a background computer use capability on par with Codex's version from May this year.
- He argues shipping great features first encourages other labs to ship too, and OpenAI will keep doing so.
- He claims GPT models effectively solved computer use in practice about four months ago, and urges the industry to invest more in training models for computer use.
- As models become more central to businesses and economies, he argues, the value of computer use only increases.
More from coding & agent
- LangChain's OpenWiki draws early praise from community — hwchase17 · 2026-09-09
- Econometrician: every agent-driven econometrics task still needs manual standard-error fixes — danielrock · 2026-09-09
- Meta's Muse AI agent can now search, compare and book travel via Duffel connector — chetanp · 2026-09-09
- OpenAI's Navier–Stokes run: train-while-deploying and 10,000 coordinated agents — DataLearnerAI · 2026-09-09
- Hermes Agent adds Perplexity Search integration, tapping an index of 400B+ URLs — Teknium · 2026-09-09
- Dev open-sources contextwise: BM25-routed MCP wrapper cuts 40k tokens to 7k — b-dub-d · 2026-09-09