Reps With Friends — Master Plan

Version 0.1 · 26 Aug 2026 · Alexei Brown

0. Strategy in one paragraph

Win the empty niche — handicapped, small-group fitness competition delivered inside the group chat — by shipping a playable WhatsApp/Slack bot fast (weeks), using in-browser pose counting for rep verification ($0, no approvals), and letting the chat broadcast loop do acquisition. The app is the scoreboard and the archive; the chat is the arena. Monetise later through corporate wellness once the loop is proven.

1. Phases

Phase 0 — Due diligence & alignment (this week)

[x] Private repo + research (OpenGym landscape, wearables) — done, see docs 04/05
[x] Business analysis — done, see doc 01
[ ] Reply to Ben → get the Figma + full blueprint (ask explicitly)
[ ] Call: align on MVP scope, equity conversation, roles. Key questions below.
[ ] Audit the Figma against this plan; flag what's missing/rubbish/needs room (he asked for it)

Questions for the call:

Handicap mechanics as he imagines them — self-assessed baseline? Historical? (Our v1: self-reported fitness tier → effort multiplier; v2: %HRR-based.)
Rep verification expectations — is camera-in-browser acceptable for MVP?
Charity pot: who holds the money? (Platform vs. honour-system vs. partner charity)
Corporate mode: any live prospects/LOIs? That changes build priority.
Equity: what split is he imagining, vesting, IP assignment on the blueprint.
Who else is involved? (Nico's role, any other collaborators.)

Phase 2 — MVP: "the 300 loop" (~6–8 weeks)

Goal: 3–5 real group chats playing weekly.

Scope:

game-core (this repo, started): match engine, handicap scoring v1, standings, charity pot ledger. Pure TypeScript, fully tested.
WhatsApp bot via Qalarc Hub (dev/dogfood) → WhatsApp Business Cloud API (prod). Commands: create match, pick exercises, log reps, standings, winner.
Slack bot (Bolt.js): same command surface, slash commands + interactive buttons. Slack first for corporate pilots.
Web app (PWA): join link from chat → log reps with camera pose counting (MoveNet/BlazePose in-browser) or manual entry → live match view → shareable result card.
Handicap v1: fitness tier (self-set, adjustable) → effort multiplier. Honest, simple, tunable. Ship it, learn, replace with %HRR in Phase 3.

Explicitly OUT: native apps, Apple Messages, cash anything, custom ML models.

Phase 3 — Effort truth & wearables (~+8 weeks)

Handicap v2: %HRR (Karvonen) — score = reps × (your %HRR ÷ your baseline %HRR for that exercise). Any BLE chest strap via Web Bluetooth; Apple Watch via HealthKit live HR; Android via Health Connect (sync-only, cross-check).
WatchOS/Wear OS companions (rep counting from wrist accelerometer).
Season/league structure (the long-arc retention fix).
Shareable clip generation (auto-cut mid-set footage + result card overlay).

Phase 4 — Business layer

Corporate mode: Slack enterprise install, org dashboards, team-vs-team leagues, invoicing. This is the revenue engine.
Charity pot mechanics with a partner charity (legal structure first).
WHOOP/Garmin/Google Health cloud verification (recovery-adjusted handicaps).
Messenger bot; revisit Apple when/unless an API ever appears.

2. Architecture (summary — full detail in doc 03)

┌─────────────┐   ┌──────────────┐   ┌─────────────┐
│ WhatsApp bot│   │  Slack bot   │   │   Web PWA   │
│ (Hub→Cloud) │   │   (Bolt)     │   │ (React)     │
└──────┬──────┘   └──────┬───────┘   └──────┬──────┘
       └────────┬────────┴─────────────────┘
                ▼
        ┌───────────────┐     ┌──────────────┐
        │  RWF API      │────▶│  Postgres    │
        │  (Hono/TS)    │     │  + Redis     │
        └───────┬───────┘     └──────────────┘
                │
        ┌───────▼────────┐
        │  game-core     │  ← pure TS library: match engine,
        │  (this repo)   │    handicap scoring, pot ledger
        └────────────────┘

Monorepo (Bun workspaces): packages/game-core, apps/api, <code>apps/bot-whatsapp</code>, <code>apps/bot-slack</code>, <code>apps/web</code>.
Chat adapters are thin — they translate platform events into the same command bus, so adding Messenger later is one adapter, not a rewrite.
Match state in Redis (live standings), history in Postgres.
Pose counting runs client-side in the browser (MoveNet via TF.js) — no video leaves the device, no GPU bill, privacy story for corporates.

3. KPIs

Phase	Metric	Target
MVP	Groups activated	3–5 real chats
MVP	Matches/group/week	≥1
MVP	% reps camera-verified	≥40%
Growth	New users per match broadcast (viral k)	≥0.3
Growth	W4 retention of a group (still playing)	≥60%
Business	Corporate pilots	2 by Phase 4

4. Coordination note

Another agent is researching this opportunity in parallel — reconcile findings before the call. This repo is the canonical home for all RWF work product.