001tmf/harness-forge

** 자체 Meta-Harness로 Claude Code를 끄십시오. 고정 모델, 기본적으로 비계를 진화합니다.**

-555)

마구 Forge는 끝을 달리는 Claude 부호 기술입니다 ** harness-optimization 루프 ** - propose → 점수 → Pareto-best → 반복 유지 - 개선 code around 고정 모델: 그것의 기억, retrieval, 컨텍스트 건설, 요약, 신속한 템플릿 및 도구 선택 논리. 모델은 결코 변경되지 않습니다; 비계는 더 나은 얻을.

에서 메서드의 native reimplementation입니다. Meta-Harness: 모델 하네스의 최종 최적화 (Lee, Nair, 장, 이, Khattab & Finn, 2026). 이름 * 참고 repo는 Python의 ~1,260 줄을 발송합니다. (claude_wrapper.py] + meta_harness.py) 그 직업은 drive the headless Claude: spawn a 세션, 출력, 트랙 도구 호출, 모든 로그, 루프를 파. ** Claude 코드, 그 런타임은 이미 일류 도구로 존재합니다.** So 마구 Forge는 무능한 도메인만 유지합니다 논리 — 싼 scorer — 그리고 원동적 관현으로 전체 외부 루프를 표현합니다. 전체장편 검색은 ~75 라인 대신 ~ 1,260입니다.

---

한 그림에 있는 아이디어

proposer는 mutation 연산자입니다.

Harness Forge is a Claude Code skill that runs an end-to-end harness-optimization loop — propose → score → keep the Pareto-best → repeat — to improve the code around a fixed model: its memory, retrieval, context construction, summarization, prompt templates, and tool-selection logic. The model never changes; the scaffolding gets better.

It is a native reimplementation of the method in Meta-Harness: End-to-End Optimization of Model Harnesses (Lee, Nair, Zhang, Lee, Khattab & Finn, 2026). The original reference repo ships ~1,260 lines of Python (claude_wrapper.py + meta_harness.py) whose job is to drive a headless Claude: spawn a session, parse its output, track tool calls, log everything, loop. Inside Claude Code, that runtime already exists as first-class tools. So Harness Forge keeps only the irreducible domain logic — a cheap scorer — and expresses the entire outer loop as native orchestration. The whole search becomes ~75 lines instead of ~1,260.

---

The idea in one picture

seed the frontier with the incumbent harness (the thing to beat)
repeat:
    PROPOSE   k candidate harness variants     ← parallel proposer agents write code
    VALIDATE  each imports / type-checks
    SCORE     each on a held-out-protected eval ← a $0, deterministic scorer
    FRONTIER  Pareto-merge: quality up, cost down, floor-respecting
final: score the frontier once on the untouched test split

The proposer is the mutation operator. The frontier is the search memory. The model is frozen throughout — which is exactly why this fits a fixed / off-the-shelf-API deployment, where you can’t change the weights and the gain has to come from the harness.

The paper’s headline result was +7.7 accuracy points at ~4× fewer context tokens on text classification — a pure harness-side win. Harness Forge reproduces that shape of result natively.

---

Why native?

claude_wrapper.py is a hand-rolled agent runtime. Claude Code is an agent runtime. So every orchestration piece has a native equivalent, and the Python driver becomes redundant: