Relic turns multi-agent collaboration failures into executable org protocols, +5.7pp delivery
ulab-ai · hf · 2026-09-29
The Relic paper converts recurring multi-agent collaboration failures into organization-owned, executable, revisable protocols.
Motivation: one coding agent changes an interface while another builds on the old version; a conversation fixes the episode, but how does the lesson persist when members change?
Mechanism
- Members reflect on visible work, propose rules, and govern adoption
- Adopted protocols bind triggers, responsibilities, evidence, and consequences at runtime, remaining revisable or retireable
- Example: repeated integration friction yields an interface-review rule governing later PRs
Results
- 360 controlled runs, 10 software workloads, 3 models: complete-contract delivery 14.06% → 19.76% (+5.71pp)
- Fresh-member transfer: 25.4% no protocol, 34.6% text rules, 41.2% executable bindings
- Best reported on CooperBench (367/477, 76.9%) among peer-structured systems; beats Solo on the fixed same-model subset
Core finding: collaboration experience can become persistent organizational state outliving its creators.
More from coding & agent
- RSI Arena: 8 AI agents get 1,000 GPU-hours each to train a better model live — my_cat_can_code · 2026-09-29
- One-shot music video: Opus 5.5 + Runway MCP produces 'Words into Worlds' — tlakomy · 2026-09-29
- Anthropic breaks down Claude effort levels: when Max is overkill and when it pays off — rubenhassid · 2026-09-29
- MIT researcher: Opus 5.5 built an interactive JS slide deck for a 15-minute talk — davidbau · 2026-09-29
- Open-source VeriChron AI rebuilds past compliance states with bitemporal data, knowledge graphs and agents — LoksaiPalyam · 2026-09-29
- An ADR prompt for AI: how Staff+ engineers make architecture decisions — JafarNajafov · 2026-09-29