A conversational AI turns a rough bug report into a coding-agent investigation
At a glance
The user puts scattered context through ChatGPT or Copilot before handing a bounded task to a coding agent. This workflow is listed under coding workflow, debugging, agent handoff.
- Agents / tools
- ChatGPT, Codex, GitHub Copilot, Amazon Q Developer
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Gather the broader evidence for a bug, including logs, personal notes, and information outside the code repository. 2) Discuss the problem with ChatGPT or Copilot and ask it to turn the rough goal into a specific investigation request and verification steps. 3) Give that request to Codex for a personal project or Amazon Q Developer at work. 4) Bring the coding agent's findings or changes back to the conversational AI for review and follow-up discussion, then decide whether the result answers the original issue.
How it went
The author used the loop with ChatGPT and Codex on a personal project and with Copilot and Amazon Q Developer at work. They found it especially useful for debugging when relevant clues were spread across sources.
Worth knowing
The post reports the author's experience without a measured time or quality comparison. The conversational model helps frame and review work; the coding agent still performs the repository investigation.
