china-qijizhifeng/agentic-harness-engineering — GitHub trend

FINDING #1569 · UNIT ID 1218737406

china-qijizhifeng/agentic-harness-engineering

Official AHE code — Agentic Harness Engineering: observability-driven automatic evolution of coding-agent harnesses (concurrent w/ meta-harness). NexAU-AHE reaches 84.7% ± 2.1 pass@1 on Terminal-Bench 2 (GPT-5.5). Lifts GPT-5.4 69.7→77.0% over 10 iters, beats Codex/ACE/Training-Free GRPO; frozen harness transfers to SWE-bench-Verified.

[ PYTHON ][ ORG ][ GITHUB ↗ ]

SURPRISE SCORE

0.00

Score Breakdown

SURPRISE0.0205

ENGAGEMENT0.20

FRESHNESS1.00

SCORE = SURPRISE × ENGAGEMENT^0.7 × FRESHNESS × VISIBILITY × CONFIDENCE

SURPRISE = WINDOW STARS / DAYS / (AUDIENCE + FLOOR)

5% OF STARS IN ARCHIVE

Growth Telemetry

VELOCITY /D

5.14

ACCEL

-0.64

RETENTION

33.3%

PEAK 2026-07-09 · FORK-RETENTION 0.0% · 36 STARS / WINDOW

Author Audience

AUDIENCE

211

FOLLOWERS

OWNER ★

754

Engagement Signals

FORKS

ISSUE AUTH

PR AUTH

UNIQUE STARGAZERS 36 / 36 (DIVERSITY 1.00)

Why This Is A Finding

china-qijizhifeng/agentic-harness-engineering собрал 36 звёзд за окно, тогда как у автора всего 30 подписчиков — эффективная аудитория ≈ 211. Это даёт surprise-индекс 0.0205 (звёзды относительно охвата автора, а не в абсолюте). Удержание форков 0.0% и 0 внешних контрибьюторов отделяют реальный инструмент от разовой вспышки. Акселерация отрицательная — внимание остывает после пика.

Related Findings

RANKS ABOVE 77% OF 6892 FINDINGS

#01MDX-Tom/gpt-5.6-instruct[ PYTHON ]0.86 #02Git-creat7/grokRegister-cpa[ PYTHON ]0.67 #04William-Lu-stack/LuxyAI[ PYTHON ]0.52 #05AlephAITech/WorkBuddyGuide[ PYTHON ]0.51 #06Cjbuilds/Codex-Orchestration[ PYTHON ]0.51

METRICS IN CONTEXT

MEDIAN ACROSS ALL 6892 FINDINGS · Δ vs MEDIAN · PERCENTILE = SHARE RANKED BELOW

METRICVALUEMEDIANΔ MEDPERCENTILE

SCORE0.010.00+0.00ABOVE 77%

VELOCITY5.144.14+1.00ABOVE 56%

RETENTION33.3%40.6%-7.3 PPABOVE 37%

FORKS8089-9ABOVE 47%

SURPRISE0.020.01+0.01ABOVE 68%