Strategy

Strategic decisions.

Ratified directional decisions and open forks. Each entry is backed by a decision record in decisions/; the evidence behind them lives in experiments/ and research/.

ratified filed 2026-07-16

Solo-first, local-first — SaaS out of scope

The analysis substrate runs on operator hardware; footage and event data never require a hosted service. Two portability seams stay open (files + SQLite timeline with JSON export; headless CLI pipeline). Revisit trigger: pipeline proven AND a second paying operator exists. decisions/0001-solo-first-local-first.md

ratified filed 2026-07-16

Photography repo as internal reference — four adoptions, one divergence

Adopted from the photography pipeline: sightings-then-human-resolution identity, single-source taxonomy with drift check, content-hash dedupe, versioned extraction provenance. Deliberately diverged on per-clip operator review (their product is findability; this one IS the editorial pick). decisions/0002-photography-repo-reference.md

ratified filed 2026-07-16

Stage 6 deploy: provenance portal published, product stays local

Operator call, 2026-07-16: this portal — research, storyboard, experiment provenance — ships to a public URL (Cloudflare Pages). The product itself remains local per ADR-0001; footage, the event timeline, and the review queue never leave operator hardware. The Try embed goes live only when the local queue is running.

When something lands here

A decision record is filed when a question can't be settled by reading code or existing records alone — delivery shape, scope boundaries, build-vs-buy. Every entry cites its evidence: measured experiment results on the operator's real footage, never projections.