Active Work and Destroying Context
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.4k
- PR merge metrics
- PR metrics pending
Description
What version of the Codex App are you using (From “About Codex” dialog)?
26.820.60940
What subscription do you have?
Plus
What platform is your computer?
Windows
What issue are you seeing?
Codex 5‑Hour Usage Limit Triggering After 1–2 Prompts, Interrupting
I am experiencing severe problems with the Codex 5‑hour usage limit. The limit is being consumed extremely quickly — often after only one or two prompts — even when the workload is minimal. This makes Codex effectively unusable for any real coding workflow.
Observed behavior:
The 5‑hour quota drains abnormally fast, sometimes within a few minutes of light usage.
I frequently hit the limit after only 1–2 prompts.
Active turns are interrupted immediately when the quota reaches 0%, even if the turn was already accepted and running.
After interruption, Codex performs extensive recovery work (reconstructing workspace, reloading context, validating partial edits, rerunning tests), which consumes additional quota.
This creates a feedback loop where Codex spends most of its time on recovery instead of forward progress.
Context is often lost entirely after the interruption, even in very short sessions.
The agent has no reliable signal about remaining quota and cannot checkpoint safely.
The effective usable time is far below 5 hours — often just a few minutes.
Impact:
I cannot complete even small coding tasks.
The agent repeatedly wastes quota on recovery cycles instead of productive work.
The system becomes unusable for iterative development, debugging, or multi‑file tasks.
The experience is significantly worse than previous Codex behavior.
Expected behavior:
Any of the following would resolve the issue:
Allow already‑accepted turns to finish even if the quota reaches 0%, while blocking new turns.
Suspend active work at the quota boundary and resume it cleanly after reset, without requiring full workspace reconstruction.
Provide the agent with a reliable remaining‑quota signal so it can checkpoint safely before interruption.
I am requesting removal of the 5‑hour limit. It is absolutely horrible.
What steps can reproduce the bug?
just using it, preformance is now horrible
What is the expected behavior?
No response
Additional information
No response
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reproducing the quota drain and active-turn interruption on Windows with one or two minimal prompts, then observe whether context is lost and recovery work begins. Done means the quota behavior and interruption handling are corrected so accepted work finishes or resumes without destructive context reconstruction.
Written by the indexing model from the issue text.
Assessment
- Domain
- ai
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100