Its theme's rank, 14 days: 5 on the Claude Code sheet
4 people reported

All time, since Sep 23.

Durations not from timestamps

“Fabricated or impressionistic quantities (durations and effort estimates) stated as fact”

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/claude-code/issues/1231/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 in each of the last two weeks. Not fixed, but a workaround is known.

Durations and time labels stated from impression, not timestamps. Durations, effort estimates and relative-time phrases like "yesterday" or "just now" are asserted without computing them, even when timestamps are in tool output.

Is this your bug?

File my report
Where
Model responses
Versions
2.1.222 to 2.1.310
Platforms
Windows
How bad
Breaks a normal workflow.

Part of: Ignores instructions and rules46 bugs

  1. Unauthorized edits despite rules14 people
  2. Ignores brevity instructions8 people
  3. Agent ignores instructions17 people
  4. Acts without authorization and ignores stop interrupts8 people
  5. Ignores claude.md rules and skills8 people
  6. Memory and protocol files not reloaded after context compaction1 person
  7. Saved memory not applied8 people
  8. Verdict reversal under pushback1 person
  9. Ignores user instructions5 people
  10. Stale schedule lane rows1 person
Show all 45Show only the largest 10
  1. Unverified search claims stated1 person
  2. Replies exceed brevity rule2 people
  3. Model expands task scope5 people
  4. Writes temp files against rule1 person
  5. CLAUDE.md rules ignored3 people
  6. Ignores provided source files4 people
  7. Banned .txt memory dumps1 person
  8. Ignores explicit user instructions3 people
  9. Model ignores CLAUDE.md rules, skills and enforcement hooks3 people
  10. Short options despite long-form rule1 person
  11. Git run despite git ban1 person
  12. Ignores instruction to research first3 people
  13. Task list not advancing3 people
  14. Guesses on ambiguous instructions1 person
  15. Delays filing requested reports1 person
  16. Direct user questions ignored3 people
  17. Agent acts on stale facts1 person
  18. Asks questions instead of building1 person
  19. Skips requested actions on computer2 people
  20. Claims deploy tool removes old links without verifying it1 person
  21. Ssh commands run in zsh1 person
  22. Commits leave stale documentation1 person
  23. Written rules ignored intermittently1 person
  24. Ignores absolute file path rule1 person
  25. Remote host builds not checked1 person
  26. Issue created with empty body1 person
  27. Stale document dates in commits1 person
  28. Skips design approval2 people
  29. Issues missing documents table1 person
  30. Bundles several questions in one1 person
  31. Unasked messages to other sessions1 person
  32. Ignores CLAUDE.md rules2 people
  33. Agent reports tasks done without verifying against real systems2 people
  34. Detours before filing issues1 person
  35. Opus 5.5 ignores loaded rules19 people

Not quite your bug?

  1. Guessed timestamps in records2 people
  2. Unverified guesses reported as fact27 people

Original reports4, newest first

  1. Claude consistently overestimates its own task durations#107727
  2. States durations of its own work and effort estimates from impression, not from the timestamps it has (hours claimed for minutes; estimates 20 to 50x off)#105469
  3. [feedback] Model: timestamps get invented from elapsed work; compaction summaries carry them forward as facts#97775
  4. Relative-time phrases ("yesterday", "just now") derived from conversation distance instead of timestamps (successor to #79731)#96325

Do you work on Claude Code?

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