GitHub · by Tamara Tran

fast-jev-compaction

Claude Code compaction that deletes instead of summarising.

Alex Volkov

@altryne

This is actually insane. This uses @typesafeai Jev model, as a plugin in Claude to review all the un-nesseasary tool calls, and it takes 1s to run! Like, literally, 1 second to take my Claude session from nearly 1M to ... 86K tokens! 😮 Ask your claude to install it and be amazed Use this prompt ``` Install, and configure : https://github.com/tamaratran/fast-jev-compaction ```

Sep 18, 2026 · 8.4K likesOpen on X

Tamara Tran’s Claude Code plugin replaces the compaction summary with Jev decisions. Every tool call and every tool result is scored in one fast request, the stale ones are dropped or truncated, and everything kept stays word for word. Nothing is ever rewritten.

Most compaction asks a language model to summarise the old turns, and a summary is lossy: a file path, an exact error, a constraint or a command can disappear even when it matters later. This plugin never writes a new sentence. It only deletes tool calls and tool results that Jev says are no longer needed, and it asks Jev while showing it the whole conversation. User and assistant text stays verbatim and in order.

The state sent to Jev is the conversation so far, oldest first, with every tool result replaced by a short note such as “ok, 4213 chars (omitted)”. That state is fitted into a 25,000-token budget in stages, each applied only if the last was not enough: tool inputs truncated to 1,000, then 200, then 60 characters; long texts abridged to head and tail; old messages collapsed to a note; old tool calls reduced to one line each. If it still does not fit, compaction throws rather than guessing.

For every call that is not pinned, Jev is asked two yes-or-no questions: should the call itself stay, and should its result stay verbatim. Calls in the first message and in the newest messages are pinned and never touched. Questions are split across as many requests as needed to stay under Jev’s 32k request limit. Alex Volkov reported a Claude session going from nearly a million tokens to 86,000 in about one second. Theo has published the case against it: lost reasoning, higher cache-write costs, and agents that loop.