Benchmark Audit

  1. Headroom at 77,253 Stars: 300 Log Lines Compressed to Zero, and Four Different Savings Numbers for the Same Tool

    Headroom advertises 20% fewer tokens for coding agents and 60-95% for JSON. Measured on 2026-10-22, a 300-line log became 35 tokens: '[300 lines compressed to 0. Retrieve …

    AI

  2. Ponytail at 169,651 Stars: a 2,721-Byte Ruleset, and the 26% Saving That Shrinks to 1.7% Once the Code Is Checked

    Ponytail sells a -53% code and -26% cost win from a single markdown file. Recomputing its own published per-task table by tier: on the 18 tasks with execution-based …

    AI