A model spends 17 hours rewriting its own harness for eleven points

Claim

Kimi K3 spent 17 hours rewriting its own Cline harness to jump from 77.5% to 88.8% on Terminal Bench while cutting costs.

Reported metric: 77.5% to 88.8%

Reporting date: 2026-07-31 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to July 31, 2026 (theinnermostloop.substack.com)