Shanghai AI Lab's Harness-of-Harness Boosts Long-Horizon Coding Agents by 52%
Shanghai AI Lab's Harness-of-Harness framework adds continual planning-development-testing iteration on top of coding agents, enabling multi-day autonomous software development. It gains an average 52% across three benchmarks (71.5 vs 58.2) and autonomously built a playable FPS game.
2026-09-05 ~ 2026-09-06 · 4 related posts
- Shanghai AI Lab's Harness-of-Harness runs 70+ iterations to build an FPS game autonomously — aigclink · 2026-09-05
- Harness-of-Harness paper: 52.25% avg gain across 3 benchmarks, builds playable FPS unattended — aigclink · 2026-09-05
- Harness-of-Harness carries state across runs: 71.52 vs 58.24 on long-horizon coding — rohanpaul_ai · 2026-09-06
- Harness-of-Harness framework boosts autonomous coding agents by 52% over multi-day runs — rohanpaul_ai · 2026-09-06