multi-agent-coordination

2026-09-24T02:23:45Z: concord-parallel-coding-agent-coordination closed (absorbed) — Scott already holds and implements this coordination pattern in Paperclip and his deterministic agent-control-plane work: explicit task ownership, durable shared state, status signals, and bounded parallel workers.
2026-08-19T23:38:23Z: cumora-multi-agent-coding-harness closed (faded) — Cumora is another unvalidated multi-agent coding harness in a territory Scott already treats as a systems-and-verification problem in “Micro-Agents Architecture,” “Shared Blackboard,” and “How to Do a Month’s Work in 1 Day.” Th
2026-08-19T16:50:30Z: slivingdoc-git-s3-agent-memory closed (faded) — Slivingdoc restates architecture already held in Scott’s Shared Blackboard, Markdown OS, Long-Running Agents, and Durability Beats Coordination pages: replaceable agents coordinate through durable, versioned external state. Its Gi
2026-08-14T18:37:35Z: parley-cross-team-agent-coordination closed (faded) — Parley’s stated design converges with Scott’s model-to-model delegation, explicit handoff contracts, and auditable agent control-plane work, potentially providing an external comparison for Paperclip. Relevance is medium bec
2026-08-11T17:41:59Z: protolink-replayable-agent-jury closed (faded) — Protolink independently packages Scott’s existing replay, agent-observability, and multi-agent deliberation ideas into a jury-style environment that could test whether influence and decision provenance are reconstructable in prac
2026-08-09T21:31:24Z: agenthelm-versioned-agent-memory closed (faded) — Scott already treats durable, versioned external state as the coordination substrate for concurrent agents and implements adjacent wiki/MCP infrastructure in dev_wiki. The supplied evidence adds no test results showing AgentHelm
2026-08-09T19:41:22Z: open-ended-agent-coordination-benchmark closed (faded) — The reported ablation provisionally challenges Scott’s load-bearing claim that durable external state, rather than better inter-agent coordination, is what makes multi-agent loops reliable. If independent follow-ups confi
2026-08-09T10:31:58Z: claude-code-session-messaging closed (absorbed) — Anthropic is productising the concurrent-session coordination pattern Scott already specified through Shared Blackboard, Micro-Agents Architecture, and parallel agents by concern, creating a strong dated-receipts opportunity. It