Adding two gates with Jev to ChatGPT's bug-fix loop
The author places Jev in front of ChatGPT to classify and screen bug reports, so only well-evidenced reports worth handling enter the fix pipeline.
YOUR CHATGPT AGENT SHOULD NOT WAKE UP FOR THE SAME UNCHANGED BUG. most coding loops send every report straight to the model. even when the issue hasn’t changed. even when another run already claimed it. even when the last attempt failed for the same reason. put two gates in front of ChatGPT: 01 / CODE same issue, same version → skip. new evidence → continue. 02 / JEV ask it to classify the new report and check whether it includes steps, the failure and expected behavior. uncertain answer → review. 03 / CHATGPT take the eligible case, reproduce the bug, make a focused fix, run checks and return the diff. Jev chooses which case deserves attention. ChatGPT still has to prove the bug exists and show that the fix works. a route is not evidence. that’s the loop I mapped out below ↓
This post quotes another post: x.com/mr_kozh/status/2103131220884529370