OpenAI Codex upgrades: GPT-5-Codex and first-class PR code review
September 2025: GPT-5-Codex launch with trained code-review behavior and GitHub auto-review / @codex review flows.
OpenAI’s Codex upgrade post puts code review beside codegen as a trained skill: intent-vs-diff checking, repo reasoning, and even running tests. GitHub integration supports automatic reviews and @codex review mentions.
Why it matters for code review
Frontier labs treating review as a first-class trained behavior (not a prompt wrapper) raise the floor for everyone else. Also note the conflict-of-interest pattern again: same agent family writes and reviews.