Microsoft's Self-HPO: Coding Agent Redesigns Its Own Harness, +7 Points on SWE-bench
mark_k · x · 2026-10-21
Microsoft's Self-HPO has a coding agent redesign the software it runs inside. The agent studies its own failed attempts and proposes changes to its "harness" — the prompts and runtime code controlling how it uses tools. The best changes are selected, the model receives RL inside the improved setup, and remaining failures become the next round's focus.
Using Qwen3.5-35B-A3B, average success rises from 57.8% to 64.8% across SWE-bench Verified, Multilingual and Pro. Improving model and harness together beat optimizing either alone, and since the agent itself proposes harness revisions, no separate stronger model is needed.
Code, trained checkpoints and resulting harnesses are publicly available.
Related event: Microsoft's Self-HPO Lets Coding Agents Rewrite Their Own Harness(2 posts)→
More from coding & agent
- Developer builds himself a 'Rork MoTS' agent on Grok Bot, aiming for full autonomy before GTA 6 — rudrank · 2026-10-21
- OpenAI Live API tip: control voice sessions from your backend via sideband WebSocket — pbbakkum · 2026-10-21
- OpenAI's MultiLangBench: PHP scores highest (91.3%) at building a database engine from scratch — ricklamers · 2026-10-21
- UserProof lets Claude Code watch your user session recordings and boost conversion for $29/mo — doooyle · 2026-10-21
- dphnAI's distributed inference network served 32B DeepSeek tokens in 24h, ~6% of Venice's daily volume — Ronangmi · 2026-10-21
- Superconductor revamps agent picker spanning 8 coding agents with search and favorites — sergeykarayev · 2026-10-21