H
Harness2
Google research framework in which an editor model rewrites an agent harness between runs of a task and carries lessons to the next task.
0 members list Harness2, unchanged over the last 12 weeks, data as of 21 October 2026
Usage on Stackness
0membersunchanged over the last 12 weeks
Members who list Harness2 in their Stack, by week.
Be the first to add Harness2
Nobody on Stackness runs Harness2 yet. Add it to your stack and you're the first member running it.
Nobody on Stackness runs Harness2 yet.
Sign up and add itPosts about Harness2
- What is a self-improving agent harness? Google's Harness2, and whether tuning pays
A self-improving harness rewrites its own prompts, skills and hooks between runs. Harness2 lifted GLM-5 11.5 points, with Opus 4.8 editing, October 2026.