Writing

From Swarm to Protocol, Part 1 of 4 — the common arc: one-shot vibes, plan gates, a sixteen-agent swarm, and the discovery that the bottleneck was the human switchboard.

From Swarm to Protocol, Part 2 of 4 — killing the swarm for a cross-model agent pair, meeting correlated errors, and learning to parallelize reads, not writes.

From Swarm to Protocol, Part 3 of 4 — recursing the pair into an orchestrator team, the delegation failures that followed, and the three rules that held it together.

From Swarm to Protocol, Part 4 of 4 — what the machinery costs, what the floor actually catches, and why stripped of the agents it was just engineering all along.

My first blog post on this site.