$ Farpost · Build Plan

Port the two pieces of domain logic that already work in production — reputation-ranked dispatch (see Farpost Dispatch) and NFC-tagged building records with per-category fact staleness — onto the converged siloes stack (Fastify, Drizzle, Postgres). This phase is deliberately a port, not a rewrite: the logic is already proven, only the runtime underneath it changes.

Replace the original stack’s ad hoc auth with better-auth’s passwordless magic-link flow, matching Vocare’s already-proven pattern, so every siloes project shares one design even though each keeps its own separate accounts and database.

Atlas and Pulse started as standalone portfolio demos (Atlas, Pulse) because there wasn’t time to build them into the real product on its own stack. Now that the product itself is being rebuilt, both ideas move from standalone demo to real feature: Atlas’s rural-density map overlay and Pulse’s AI-assisted coaching tips get built against real dispatch and building-record data, not the illustrative data their demo pages use today.

Once the rebuild covers the original system’s real usage, cut real traffic over from the MongoDB/FastAPI/Twilio production system to the new build, and retire the original stack. No date is set for this yet — it depends on how Phases 1 and 2 actually go, not a target picked in advance.