Unclear It is not known whether the mode that enables the faster resume is on by default.
What
A session's transcript is the saved record of a conversation, which Claude Code reads when you resume it. Claude Code can now write and read a side file called resume-head.jsonl to make that loading faster.
- When resuming, it loads
resume-head.jsonland checks it. If the file fails that check, it reads the whole transcript as before. - It can work out and write a new
resume-head.jsonlto speed up later loads. If that fails, it logs "could not work out a new resume-head.jsonl". - The faster path skips the part of the history from before compaction (the point where earlier conversation was summarised to save space). Setting the environment variable
CLAUDE_CODE_DISABLE_PRECOMPACT_SKIPturns that skip off.
The skip only runs when an internal mode is set to "on".
Why
This may make resuming large sessions noticeably quicker. If resuming ever looks wrong, CLAUDE_CODE_DISABLE_PRECOMPACT_SKIP makes Claude Code read the history without the skip.
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubt
It is not known whether the mode that enables the faster resume is on by default.