Microsoft open-sources SkillOpt: training skill docs lifts GPT-5.5 accuracy 23.5 points
sanjaykalra · x · 2026-09-27
Microsoft Research open-sourced SkillOpt, which turns the most valuable asset of an enterprise AI program into a markdown file. Instead of touching model weights, it scores agent runs and iteratively edits the skill document the agent reads, keeping an edit only if it beats a held-out validation set. On GPT-5.5, Microsoft reports a 300–2,000 token trained skill file lifted accuracy by 23.5 points in direct chat and 19.1 points inside Claude Code — though these are Microsoft's own benchmarks, and the author wants them validated on real enterprise workloads. Framed as "Rent the Gym. Guard the Playbook.": the model is rented, but the bestskill.md is yours, and the paper claims it transfers across models and harnesses without retraining. Your validation set becomes the moat.
More from coding & agent
- Improved chat template for Laguna XS/S 2.1 adds forced thinking and stability fixes — arbv · 2026-09-28
- Microsoft open-sources Data Formulator, an AI data analysis tool that builds editable charts from plain English — usamawahabkhan · 2026-09-28
- Opus 5.5 Builds a Fishing Game via Unity and Blender MCP — noahsolomon · 2026-09-28
- Jane Street's Yaron Minsky worried by vibe-coding codebases in languages he can't read — burny_tech · 2026-09-28
- Casita launches as Rust content-addressed store to rethink Nix for agentic coding — letandrewcook · 2026-09-28
- Applied AI hiring: weeks of agent tinkering can beat years of experience — brandon_galang · 2026-09-28