Skip to content

ROADMAP.adoc

Latest commit

 

History

History
78 lines (62 loc) · 3.63 KB

ROADMAP.adoc

File metadata and controls

78 lines (62 loc) · 3.63 KB
Raw

flat-mate Roadmap

Current State

flat-mate v1 is Squad Audit: an already-formed student group gets a harmony + feasibility audit, an auto-drafted house agreement, and a search brief to take to the listings market. flat-mate sits upstream of SpareRoom and hosts no listings. Authoritative spec: docs/design/squad-audit-v1.adoc (where code and that document disagree, the document wins).

Engine v2 is a tested library at packages/shared/src/matching/ (10 modules, 5 test files, 35 unit tests passing). The apps (apps/api, apps/web, apps/mobile) are not yet wired to it; the legacy v0 swipe+listings stack remains in-tree marked LEGACY pending a quarantine decision.

Target wave: Jan–May 2027 London second-year wave; one university to seed; legal posture resolved by December 2026.

What we never promise: a "stable matching" — the roommate market is non-bipartite, stable matchings typically do not exist at scale (Chin & Michelen 2026), couples make existence NP-complete, and households of 3+ are hedonic games whose core stability is Σ₂ᵖ-complete to decide. Product copy and mechanism design must respect this.

The roadmap below is sequenced by liquidity: each stage needs more marketplace participation than the one before it.

NOW — done 2026-06-11

  • ✓ Squad Audit v1 design decision record (docs/design/squad-audit-v1.adoc)

  • ✓ Matching engine v2 (packages/shared/src/matching/): directional behaviour/tolerance scoring, feasibility/harmony split with banded verdicts, auditGroup (pairwise matrix, budget intersection, dispersion detection, shared-space minimum forecast), conceal lattice T0/T1/T2 with effectiveTier = min, agreement + brief generators; 35 tests

  • ✓ proven-epistemic reference core — the conceal/disclosure semantics mirror the formally verified Idris2 core in proven-servers (ADR: proven-servers/docs/decisions/0002-add-proven-epistemic-disclosure-core.md)

NEXT

  • ❏ Wire apps/web to auditGroup — audit-first UI (the apps currently do not call engine v2 at all)

  • ❏ Legacy v0 quarantine decision — swipe+listings stack (domain.js/encoding.js, apps/api, apps/web, apps/mobile): quarantine, rewrite, or remove

  • ❏ Privacy notice + retention schedule + verified VerisimDB erasure and subject-access-export story (Art. 25 obligations named in the design record)

  • ❏ ICO controller registration decision

THEN

  • ❏ Solo top-up discovery for groups with open slots — conceal tiers (built in v1) plus a consent lifecycle (Proposal → mutual accept → Match), restoring the mutuality property the v0 swipe domain had

  • ❏ DA rounds for solo-joins-household — the one theory-clean centralized mechanism (genuinely bipartite); one vacancy per household per round; couples via heuristic + blocking-pair verification, never a guarantee path

LATER

  • ❏ Calibration flywheel — anonymised outcome telemetry (moved in; 1-month / 3-month satisfaction); literature-seeded weights become fitted; one learning cycle per academic year

  • ❏ Blocking-pair stability audit over the discovery graph (advisory, private, bounded to pair deviations)

  • ❏ Borough/commute prefilters in candidate generation when discovery lands (London-only invariant stands; the audit flow is location-light)

Estate invariants (unchanged)

VerisimDB sole persistence; London-first; Deno JS-first with the four-file TS carve-out in apps/api (see .claude/CLAUDE.md); MPL-2.0; mobile via Expo.