ChatGPT's Codex just got a huge upgrade that makes it more powerful than ever - what's new (www.zdnet.com)

OpenAI has moved Codex out of research preview to "generally available," rolling out a Slack integration, a Codex SDK (plus a GitHub Action for CI/CD), and new admin/management features for Business, Edu and Enterprise customers. Developers can now call Codex programmatically from code, trigger it in automated pipelines, or @mention it in Slack; Codex runs tasks in isolated cloud containers (kept for logs and diffs) that admins can inspect or delete. Admin controls also include managed configuration overrides for the Codex CLI and IDE extension, monitoring of actions, and analytics that report usage and code-review quality. Starting Oct. 20, Codex cloud-task usage will count against customer resource allocations. The update is significant because it shifts Codex from an experimental assistant into an integrated engineering toolchain component: think automated bug triage/fix attempts, AI-assisted PRs, or embedding code generation into production workflows. Technical implications include tighter enterprise governance (container lifecycle control, safer defaults, usage analytics) but also higher risk if teams over-automate—Codex can produce useful code quickly but still hallucinates or makes logical errors without firm human oversight. The SDK/GitHub Action make recursive automation straightforward, so teams should combine these capabilities with testing, review gates, and conservative permissions to avoid unsafe or low-quality autonomous changes.