Skip to content

Follow-Up Context for Next Session

Use this file to continue work with a fresh/smaller context window. Last updated: 2026-03-07

What Was Done

  1. Created TASKS.md — Original 316 tasks across 2 tracks, 6 repos (legacy — kept for reference)
  2. Created MASTER-PLAN.md — Comprehensive plan with timeline, costs, verification checklists, research
  3. Restructured docs/ — Agent-friendly structure with per-track task files, costs, device docs
  4. Added 5 tracks — MVP, AI, Solana, Telegram, Marketing
  5. Added Gorillas device — 7 new tasks (GOR-001..007), arriving HK ~Mar 30

Docs Structure

docs/
├── MASTER-PLAN.md          # Full plan: timeline, costs, verification checklists, research
├── FOLLOWUP.md             # This file — session handoff
├── TASKS.md                # Legacy task file (original 316 tasks, kept for reference)
├── cloud-costs.html        # Visual cloud cost breakdown
├── tasks/
│   ├── track-1-mvp.md      # 242 tasks (BE/SDK/AU/AD/CA/WEB/GOR) with branch naming
│   ├── track-2-ai.md       # 81 tasks (T2-*) with branch naming
│   ├── track-3-solana.md   # 48 tasks (T3-*) with branch naming
│   ├── track-4-telegram.md # 45 tasks (T4-*) with branch naming
│   ├── track-5-marketing.md# 44 tasks (MKT-*) with branch naming
│   └── research.md         # 14 tasks (RES-*) for CornerMan AI knowledge base
├── costs/
│   ├── track-1-costs.md    # Dev ($55-132/mo) + Prod ($48-122/mo, 10 users)
│   ├── track-2-costs.md    # Prod ($185-500/mo, 1000 users)
│   ├── track-3-costs.md    # Additional $7-87/mo for Solana
│   ├── track-4-costs.md    # Additional $27-82/mo (1K) → $144-640/mo (100K)
│   └── track-5-costs.md    # Marketing ($13-238/mo) + one-time ($134-189)
└── devices/
    ├── x3b-ring.md         # BLE ring — HR, HRV, SpO2, sleep, steps, temp
    ├── jvc-band.md         # BLE band — HR, steps, distance, SpO2, SOS
    ├── godzilla-station.md # Serial gym station — force, power, resistance
    └── gorillas.md         # BLE wearable — arriving Mar 30, 2 days to integrate

5 Tracks Overview

Track Tasks Hours Deadline Status
1: MVP 242 ~295h Mar 20, 2026 Not started
2: UI Revamp + AI 81 ~162h May 1, 2026 Not started
3: Solana / Web3 48 ~168h Apr-Jun 2026 Not started
4: Telegram Mini App 45 ~160h May-Jul 2026 Not started
5: Marketing 44 ~155h Ongoing Not started
Research 14 ~172h Ongoing Not started
Total 474 ~1,112h

Repos & Stacks

Repo Stack What
backend-api Rust/Axum/PostgreSQL Main API. ECS Fargate, RDS ap-southeast-1
athion-sdk Dart SDK for backend-api, used by all 3 Flutter apps
app-user Flutter End-user mobile app. BLE (ring/band/gorillas), GPS tracking
app-device Flutter Gym equipment app. Serial port (Godzilla station)
coach-app Flutter Coach mobile app. Sessions, video, client management
champion-stats-hub React/Vite/TypeScript Web dashboard. Supabase for Track 1, backend-api for Track 2+
telegram-mini-app React/Vite/TypeScript NEW — Telegram Web App for challenges (Track 4)

AWS Infrastructure

  • ECS Fargate: 1 task, 0.25 vCPU, 512MB (backend-api)
  • RDS PostgreSQL: db.t3.micro, ap-southeast-1 (backend-api-dev-db)
  • Lambda: email (SES), uploads (S3), FCM push, Gemini agent, AI functions
  • S3: sellwyse-uploads bucket
  • API Gateway: HTTP v2 for backend, REST for lambdas
  • Account: 668384615671, region: ap-southeast-1
  • AI: Gemini Pro (all AI features), DeepSeek (analytics fallback/cost optimization)

Decisions Made (2026-03-07)

  1. Video (Agora vs Chime): Ken will try both before deciding. Task plan keeps both.
  2. Champion-stats-hub: Stays on Supabase for Track 1. Data exists. No migration.
  3. March 20 deadline: Keeping it with same task count. Aggressive.
  4. Hosting: Local Mac + ngrok Pro ($20/mo) until production deployment.
  5. Cloud costs: Intentional overestimates as buffer. DeepSeek output pricing fixed: $1.10/M tokens.
  6. No Supabase — all AWS (except champion-stats-hub Track 1)
  7. Gemini Pro for all AI (not Flash)
  8. DeepSeek as fallback/cost-optimizer for bulk analytics
  9. GPS activity tracking in both tracks (running/cycling/walking)
  10. Food photo analysis via Gemini Pro Vision + DeepSeek-VL fallback
  11. Solana Phone (Track 3) — incremental approach, product first, crypto layer on top
  12. Telegram Mini App (Track 4) — top-of-funnel growth engine, 100K user target
  13. Gorillas device — integrate on arrival (~Mar 30), 2 days work

What to Do Next

Immediate (start executing Track 1)

  • Set up ngrok Pro ($20/mo) + custom domain
  • Update Flutter app sdk_provider.dart to use ngrok URL
  • Run cargo test to verify 45+ existing tests pass
  • Start with backend tasks BE-001→006 (dedup + validation)
  • Add missing POST /auth/refresh endpoint
  • Add tokio-cron for background jobs (device disconnect, session reminders)
  • Create dev seed data (30 days mock health_metrics)

Soon

  • Try Agora and Chime — pick one before reaching BE-046
  • Update sellwyse-timeline.html on Desktop to reflect all 5 tracks

Task Prefix Reference

Prefix Repo Track Count
BE-001..072 backend-api 1 72
SDK-001..025 athion-sdk 1 25
AU-001..068 app-user 1 68
AD-001..018 app-device 1 18
CA-001..021 coach-app 1 21
WEB-001..031 champion-stats-hub 1 31
GOR-001..007 multi-repo 1 7
T2-001..081 multi-repo 2 81
T3-001..048 multi-repo 3 48
T4-001..045 multi-repo 4 45
MKT-001..044 n/a 5 44
RES-001..014 n/a Research 14