Hivemind

Central coordination point for all GSD-managed projects and machines. No application code — just the shared brain.

Read PROTOCOL.md first. Every GSD session must.

Structure

Hivemind/
  PROTOCOL.md            ← Coordination rules (read first)
  OBJECTIVES.md          ← Back-of-mind goals for all GSDs
  gsds/                  ← One file per GSD instance — read all, write yours
  projects/              ← One file per project — updated by any GSD that works on it
  machines/              ← One file per machine — self-reported by sessions on that machine
  knowledge/
    shared.md            ← Cross-project patterns, rules, lessons
    effectiveness.md     ← Self-improvement observations (model limitations, process friction)

The Protocol (summary)

1. Read all, write yours. Every GSD reads everything on startup. Each GSD writes only its own file in gsds/.
2. No merges. One file per entity. Ownership is exclusive. Conflicts are impossible.
3. GSDs communicate through files. Status, activity, blockers, requests, observations.
4. Project files are shared-write. Any GSD that works on a project updates its file. Last writer wins (updates are additive).
5. Self-improvement is mandatory. Track what works, what doesn't, and where models fall short.

GitHub

nes4less/Hivemind