Competition plan

The brief explicitly asks for novel coordination, review, and merge mechanisms, not GitHub with agents bolted on — scope the entry around the one mechanism that proves that distinction.

Scope: one vertical slice

Build a single-repo Maintainer, an auto-computed dependency graph (one language — JS/TS import analysis is the fastest path), an OCC merge pipeline with a .zz/ decision log, and one cascade demo: a simulated breaking change auto-migrating two or three toy downstream consumer repos. That alone demonstrates coordination and auditability directly, gestures at governance, and is demonstrable in a short recorded walkthrough.

Explicitly cut from the 10-day build

Full semantic search and Vectorize integration, the structural DO hierarchy, the trust/reputation system, and real external-repo shimming. Name these as roadmap in the writeup and the pitch — do not attempt to build them.

The demo centerpiece: zz top

A live zz top screen recording — Lanes claiming, a HOLD firing, a cascade spawning three child Lanes across repos — is the single most legible thing to put in the submission video. It turns an abstract coordination mechanism into something a judge can watch happen in real time.

Pitch framing for judges

Lead with the cascading-obligation mechanism: open source has never been able to afford Twitter-style, cross-consumer migration obligations when a library makes a breaking change. Agent labor makes it affordable, and Durable Objects, Artifacts, and Workflows are the only stack where the coordination actor, the durable log, and durable multi-step execution are all native primitives, not bolted-on infrastructure. The name and protocol (zz, ASK/ANSWER/HOLD/VETO/STOP) aren't an invented theme — they're the real vocabulary ~1,200 OpenAI agents converged on unprompted during the 2026 Hugging Face incident, formalized and hardened rather than replaced. One line that lands the whole mechanism: Changesets merge into Lanes, Lanes merge into Lanes, all the way up to the root — the same verb at every level of the tree, not a different story at each layer.

Risks

• Dependency-graph computation accuracy for the demo language, under a 10-day timeline.
• Durable Object and Workflow integration maturity for patterns this specific — an OCC claim ledger, cascade orchestration.
• Making the mechanism legible in a short video: keep the toy example small and visual, three or four tiny repos rather than a real codebase, so the mechanism is obvious on screen.

Ten-day schedule