COOL & INSIGHTFUL: "Idle compaction" was added to Claude Code TUI in their most recent release (2.1.299).
Each Claude Code session (with at least 200k+ context) will now automatically compact after being idle for ~54 minutes, in order to refresh the 1-hour TTL and prevent the user from incurring the cost of uncached input.
This will save end users a lot of money! Although, with the way Anthropic investor day is going, maybe they should roll this change back? 🤔
Resuming a long Claude Code session after an hour used to mean paying uncached input for the whole context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →. Idle context compactionSummarizing an agent's earlier conversation to free room in the context window so a long session can keep going.Full definition → after about 54 minutes keeps the cache warm, which lowers resume costs for sessions over roughly 200k tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition →.
Terms in this piece · Glossary
context compaction — Summarizing an agent's earlier conversation to free room in the context window so a long session can keep going.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
