Student renters in London need two things at once: matching (who to live with) and listing (where to live). flat-mate handles both in one flow, with profile compatibility and listing discovery centered around London boroughs and universities.
Profiles, listings, and swipe events are stored as hexads in VeriSimDB. The API runs on Deno; the web client uses React + Vite; the mobile client uses Expo. See docs/architecture.md and TOPOLOGY.md for the full architecture.
Caveat: This is an MVP. University email validation (.ac.uk) is enforced
server-side — every profile write runs validateProfileInput
(packages/shared/src/domain.js), which rejects non-.ac.uk addresses — but it is a
domain-suffix check only: no confirmation email is sent, so address ownership is not
verified. Compatibility scoring is explicit weighted-factor arithmetic (budget ×0.4,
lifestyle ×0.4, preferences ×0.2), not a dot-product and not a trained model; the
profile vector is used only for candidate retrieval in VeriSimDB. The planned evolution
is literature-seeded weights with an outcome-calibration flywheel, not a trained model.
The flow described above is the legacy v0 marketplace; the current product design is the Squad Audit — see docs/design/squad-audit-v1.adoc.