docs: The Product Loop — nanopm's method manifesto

[nmrtn]

What

Adds METHOD.md — the opinionated methodology behind nanopm, written Shape Up / Getting Real style. Linked from the README.

Structure

• Part 1 — Diagnosis: building got fast, deciding didn't; trackers track but don't think; the unit of product work is the loop, and agents finally make its discipline affordable.
• Part 2 — Ten principles: problems before solutions · evidence is identity · every bet is falsifiable · appetite not estimates · adversarial by default · agents do / humans decide (exactly three human calls) · memory compounds · all work competes in one queue · the roadmap is an output not a document · close every loop.
• Part 3 — The Loop: SENSE → SIFT → RANK → ACT → LEARN, running on a daily/triggered tick.
• Part 4 — What we refuse: no numeric scores as fact · no estimates · no tracker rebuild · no dashboard theater · no silent automation of judgment.
• Part 5 — State of the method: honest per the method's own provenance rule — vertical chain, memory, and gates marked evidence-backed by the dogfood; the unified queue, RANK, the rendered roadmap, the tick, and LEARN backprop marked assumed / being built.

Dogfooding — the method applied to nanopm

Part 5 doesn't just claim the loop found its own gaps — it cites the receipts. Each unbuilt brick maps to a real, ranked opportunity already in nanopm's own database:

Method component	Backing opportunity (slug)
Tasks/bugs/chores as first-class work	pipeline-scoped-for-epics-not-micro-tasks (its one chosen solution: a peer task entity)
RANK — one cross-type queue → Next List	no-single-mechanism-prioritizes-across-all-work-ty
Roadmap as rendered output	roadmap-doesnt-orchestrate-into-living-plan
The tick — self-running loop	loop-runs-itself-on-a-cycle
LEARN backpropagation	opportunities-and-plans-go-stale-with-no-freshness

This PR also wires the RANK / roadmap / tick opportunities into one connected related_to cluster in the dogfood — the horizontal orchestrator the vertical chain has been waiting for. One honest asterisk, kept in the text: those opportunities carry no linked_objectives, because the current OKRs are all proof-quarter validation — none is a build-the-loop objective yet. Per the method's own rule, an empty link beats a forced one.

Provenance

Written from an audit of the 26 skills, lib/nanopm.sh, and the dogfood .nanopm/ state (42 opportunities, 35 solutions, 31 PRDs). The gap analysis: the vertical chain (signal → problem → bet → ship) is proven; the horizontal orchestrator (one cross-type ranked queue + a tick that runs it) does not exist yet — and nanopm's own discovery process had already surfaced each missing brick as a ranked opportunity before this document named them.

🤖 Generated with Claude Code