Rank, 14 days: 135 on the Codex sheet
26 people reported

All time, since Sep 28, including 18 thumbs-up on GitHub.

Premature auto-compaction

“Historical reasoning already represented in server input usage is estimated again by the client, inflating the counter used for auto-compaction.”

Is this your bug? What gets counted gets fixed. Pile on.

Paste this into your agent. It files your report, with your error and version.

I hit this bug: rapsheet.dev/sheets/codex/issues/13206/file
Follow it to file my report. Show me the text before you send.

No install. No account. Public and anonymous. The link asks your agent for one report, nothing else. What it sends

Still happening. 1 report or thumbs-up in each of the last two weeks, down from 18 a week at the mid-October peak.

Auto-compaction triggers early by double-counting historical reasoning. Client adds an estimate of prior encrypted reasoning on top of server-reported usage, so compaction starts while the indicator shows about 51% remaining.

Is this your bug?

File my report
Where
context compaction
Versions
0.159.2 to 0.166.0
Platforms
macOS, Windows
How bad
Breaks a normal workflow.

Not quite your bug?

  1. Early automatic context compaction2 people
  2. Weekly usage drains fast4 people
  3. Usage limits drain too fast4 people

Original reports8, newest first

  1. Windows desktop: premature compaction after resume; current production build lacks #57839On GitHub#59081
  2. Auto-compaction counts retained encrypted reasoning twice: internal counter = displayed + ~0.19x reasoning blob bytes (0.163.0)On GitHub#55928
  3. Premature auto-compaction in CodexOn the forum#1408178
  4. Auto-compaction while IDE shows 51% context remaining: historical reasoning estimate differs from displayed usageOn GitHub#55126
  5. Native Lite client double-counts old reasoning when x-reasoning-included is absent (54% compaction)On GitHub#53913
  6. Historical reasoning double-counted in 0.162.0-alpha.17.2, causing premature auto-compactionOn GitHub#53309
  7. Repeated early compaction across GPT-6.1 Sol and Astra: internal token counts exceed reported usageOn GitHub#49961
  8. Automatic compaction double-counts prior reasoning against reported token usage and triggers unnecessary compactionOn GitHub#49026

Do you work on Codex?

Claim its RAP Sheet to see what your users report and answer them in one place.