You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Improve how chatGPT/codex handle software engineering projects #62189
I am a retired software engineer with over 20 years of professional experience, including 22 years at BIOVIA.
I am currently a paying ChatGPT subscriber, using ChatGPT and Codex to develop real-world software applications.
I am submitting this feedback because I believe OpenAI has an excellent opportunity to improve its AI development tools for experienced software engineers.
My concerns are based on practical development experience, particularly unnecessary complexity, repeated mistakes, poor continuity and failure to follow established engineering procedures.
I would appreciate it if this feedback could be forwarded to the teams responsible for ChatGPT and Codex development.
I would also welcome an opportunity to provide further examples or participate in developer feedback discussions.
changed the title [-]Request some improvements how chatGPT/codex handle software engineering projects[/-][+]Improve how chatGPT/codex handle software engineering projects[/+]on Nov 18, 2026
For the project-continuity and engineering-procedure problems you describe, we built WorkLedger at Digital Disruptors.
The agreed specification and acceptance checks live on an assigned work item, alongside recorded coding activity, human decisions and change/test evidence. A reviewer can request changes with a note. The next supported Codex or Claude Code session picks up that task's recorded decisions, unfinished work and next action, so restarting a session doesn't mean rebuilding the project brief.
Activity
[-]Request some improvements how chatGPT/codex handle software engineering projects[/-][+]Improve how chatGPT/codex handle software engineering projects[/+]dijarvrella commentedon Nov 20, 2026
For the project-continuity and engineering-procedure problems you describe, we built WorkLedger at Digital Disruptors.
The agreed specification and acceptance checks live on an assigned work item, alongside recorded coding activity, human decisions and change/test evidence. A reviewer can request changes with a note. The next supported Codex or Claude Code session picks up that task's recorded decisions, unfinished work and next action, so restarting a session doesn't mean rebuilding the project brief.
WorkLedger provides that shared work record alongside your coding tools. Happy to demonstrate one task across two sessions through our free guided 30-day pilot, starting with one supported tool and repository:
https://workledger-ai.com/early-access?utm_source=github&utm_medium=referral&utm_campaign=workledger-engineering-continuity-2026-11-20