The Pragmatic EngineerGergely Orosz15 min readintermediate
Inside OpenAI’s agentic software factory
Summary
OpenAI has internalized Codex and ChatGPT Work as a universal coding agent, achieving >90% adoption across the company and reshaping tooling, CI/CD load, and engineering roles. The shift has forced a new agentic software factory, massive infra scaling, and a rethink of IDEs, PRs, and code reviews.
- Codex and ChatGPT Work now power almost all internal workflows at OpenAI, with >90% adoption across non‑engineering teams within months.
- Long‑running agent goals and role‑specific plugins drove rapid uptake, turning the IDE and PR model into secondary tools.
- Production load on CI/CD and version control jumped ~10× in six months, forcing continuous infra scaling and rethinking of code review.
- OpenAI built an “agentic software factory” where monitoring agents automatically spawn Codex agents to fix performance regressions.
Engineers building AI‑augmented tooling or scaling fast‑moving codebases need to see the practical impact of agentic workflows and the infrastructure challenges they create.
7/10


