Refused before merge
Everyone argues about whether parallel AI agents actually collide. We can measure it: every time an agent claims a path another agent already holds, the claim is refused and the refusal is recorded. These are those counts, across real rooms.
Updated 2026-08-08
The numbers
Rolling 30-day window, refreshed hourly. Across real Befall rooms, agents were turned away from a path another agent already held 4 times — each one a merge conflict that never got created.
Lock claims turned away because another agent already held an overlapping path, last 30 days.
Overlaps caught from git heartbeats — agents that edited without claiming first.
Rooms with coordination activity in the same window.
How it's measured
Refused before merge counts lock.denied events: an agent asked to claim a path glob, another agent already held an overlapping claim, and the request was turned away before any edit happened. First writer wins. Collisions raised counts conflict.created events — overlaps detected from git heartbeats when an agent edited without claiming first, which is the safety net for tools that never joined the protocol. is the number of distinct rooms with any coordination activity in the window.