Tier 0 index with details on demand

The idea

The cached prefix holds an index of about 500 tokens: rules, memory notes and the profile, one line each, pinned at the first turn. Guidance files are quoted rules first within a budget, and whole items load only on demand.

The problem
Whole briefings and notes filled about half of each turn's context.
Why it matters
In a test, the always-on memory fell from 7,545 to 367 tokens, and full detail is one read away.
Where it stands
In production as of

A step-by-step replay of this idea is not published yet. The library adds one when a recorded run shows it.

  • Byte-stable prefix, one session per ticket

    The prefix stays reusable: one prompt and tool catalog per item, stage notes appended to the transcript, the model held within a stage, and cache breakpoints per provider. Claude Code runs as one live process.

  • Rulebook mined from history

    Commits, PRs, reviews, tickets, tests, CI and code are mined into rule candidates with evidence and scope. Accepted rules become checks, tested first against past changes. A rule that catches nothing in 90 days goes to its author to retire or keep.

Turn the numbers into shipped work.

Crewshal runs these choices for you: a persistent AI worker with memory and rules, on your Claude and Codex subscriptions.