External Publication
Visit Post

Codex Profile Longest task does not update for completed /goal runs

OpenAI Developer Community June 7, 2026
Source

I filed a Codex profile stats bug where the Profile “Longest task” value does not update for completed long-running /goal tasks:

github.com/openai/codex

Profile "Longest task" does not update for completed long-running /goal tasks

opened 03:44PM - 07 Jun 26 UTC

      yourprism
    

bug app session

Bug Codex Profile "Longest task" is not updating for completed long-running …slash goals. This appears related to goal/profile persistence, but the key detail is that this is not only old experimental-goal history: an 8h+ /goal was run after the Profile / "Longest task" feature was already visible in the Codex app, completed successfully, and still did not update the profile stat. ## Environment - Codex CLI: codex-cli 0.130.0 - Platform: macOS - Local ~/.codex/version.json latest_version: 0.130.0 ## Details - Current Profile "Longest task" remains stuck at 3h 48m. - I ran an 8h+ Codex task after the Profile / "Longest task" feature was already visible in the app. - That task was an actual slash goal (/goal) and completed successfully. - The profile stat still did not update. - I originally remembered an older long task as roughly 41h, but local evidence shows the PRISM /goal session actually spanned about 53.63h and did not save/show in profile stats. ## Local Evidence Local Codex CLI session history exists for a task longer than my original 41h estimate: - `/.codex/sessions/2026/05/04/rollout-2026-05-04T17-27-21-019df51a-6664-7600-b1bb-c46a6f9f54c3.jsonl- Timestamp span: about 53.63h - Appears to be a PRISM/goaltask But: -~/.codex/goals_1.sqliteexists -thread_goalscurrently has 0 rows -goals_1.sqlitemigration: -1 | thread goals | 2026-05-22 18:16:08 | success=1-state_5.sqlitegoal migrations: -29 | thread goals | 2026-05-01 00:36:17 | success=1-33 | thread goal stopped statuses | 2026-05-22 18:16:08 | success=1-34 | drop thread goals | 2026-05-22 18:16:08 | success=1So the raw session history exists, but the structured goal/profile tracker did not persist the long-running goal records. ## Related Issue Possibly related to #25407, but this report is specifically about Profile "Longest task" not updating for a post-Profile, real completed/goal, plus local evidence that the goal DB has no persisted rows. ## Expected The Profile "Longest task" stat should update to the longest completed /goaltask. ## Actual The stat remains stuck at 3h 48m despite completed 8h+ and much longer goal sessions. ## Evidence Available I can provide screenshots of: - Profile showing "Longest task" stuck at 3h 48m - Completed 8h+/goal` - Longer CLI/session evidence I am not pasting raw JSONL contents publicly because session logs may contain sensitive material, but I can provide them privately through a secure support channel if needed.

The short version:

  • My Profile stat remains stuck at 3h 48m.
  • I ran a completed 8h+ /goal after the Profile / “Longest task” feature was already visible.
  • Local CLI history also shows a PRISM /goal session spanning about 53.63h.
  • The local goals_1.sqlite thread_goals table exists, but count(*) is 0.

I found the related goal-storage threads too, especially openai/codex#23984 and openai/codex#25407. The fixes/workarounds people reported there were:

  • fully restart Codex and any stale codex app-server processes so app-server and CLI versions match;
  • make sure the goals feature flag is not explicitly disabled;
  • if thread/goal/get failures persist, back up and rebuild the local Codex state DB files as OpenAI staff suggested.

That helps explain broken /goal reads/writes, but this case seems like a separate downstream persistence/backfill issue: the raw long-running session history exists, but the completed goal did not make it into the Profile “Longest task” stat.

Has anyone else seen Profile stats stay stuck after a successful long-running /goal, especially after restarting/updating Codex?

Discussion in the ATmosphere

Loading comments...