Codex checks its work for you
Codex checks its work for you
Author: OpenAI – Duration: 00:02:24
Javi walks through a logging refactor and shows why Codex self-checking is a radical change: the model runs the application, finds the correct session, and proves that the logs are still flowing. Takeaways: – Codex can validate its work by running tests and launching the application. – It excels at general refactors that affect many files. – The model can find session IDs and query tools on its own. – Verification reduces a risky manual loop in minutes. When the agent can prove you are right, you can act more quickly with less risk. Chapters: 00:00 Why Codex was a sea change 00:18 Self-check: run tests and launch the application 00:52 The task: refactor logging on many files 01:10 The risk: not breaking observability 01:28 How this was verified manually 01:35 Have the model check the logs end-to-end 01:50 It finds the session ID and queries the MCP logs 02:03 Proof: logs are still in progress, task is completed quickly
