Build Story · Wellness app · Web, iPhone, Android
Myo — Dopamine Coach: one number, one lever
Myo — Dopamine Coach puts your day's drains and restores on one balance sheet and shows the one move that would help most.

In one look
The problem
Flat mornings, scrolling on reflex, effort that feels expensive. Habit trackers count actions, screen-time limits block one app, and a detox weekend wears off.
What we built
Log your day in plain sentences and connect a WHOOP. It puts drains and restores on one balance sheet and shows one number plus the single move that would shift your baseline most today. The model is published, mechanism by mechanism, on a public science page.
Where it is now
Live on the web with a free tier and a free 60-second audit. The paid tier costs $12.99 a month or $79 a year. Native iPhone, Apple Watch and Android apps run the same engine and are in private testing, not in the app stores.
What this would have cost in 2024
About $55k and 3–4 months
2024 team: 4 people: full-stack web, iOS and watchOS, Android, designer-writer for the site and content. Built here by one founder working with AI in 28 working days over 6 months.
| Piece | 2024 hours |
|---|---|
| Web app: plain-sentence logging, Today and Week, sign-in, onboarding, AI coach with voice | ~247 |
| Scoring engine v1 to v3, calibration, forecast, risk windows, experiments, 2-year simulation QA | ~210 |
| Recovery layer: consent, protocols, taper tracking, records, recovery coach | ~147 |
| WHOOP and Apple Health sync, background jobs, staleness alerts | ~93 |
| Billing: Stripe, Apple in-app purchase, trial, coach caps | ~73 |
| Marketing site, science page, blog, Dopamine Audit with share cards | ~157 |
| Native iPhone app, Apple Watch companion, widgets, Apple Health | ~270 |
| Native Android app with Health Connect and engine parity | ~177 |
| Privacy and security pass, lifecycle emails, MCP server | ~105 |
| Total (PERT) | ~1,478 |
Method: PERT (O+4M+P)/6 per piece; low = O × $25, likely = PERT × $37.5, high = P × $50 (Accelerance 2024 Eastern-Europe band). These are estimates of what a team would have needed in 2024, not invoices.
How it was built
From first commit to current build in 28 working days over 6 months
- Day one
An early web version: log what you did and see it on a Today page with a calendar.
- The reward engine and a Recovery layer
The engine at the heart of the model, a Recovery layer behind explicit consent with its own coach, and a WHOOP connection.
- Plain sentences and a name
Log your day in plain sentences, a daily check-in and voice input for the coach. The product took its current name and moved to myo.coach.
- A price, a site and the model in public
A free tier and a paid tier with Stripe, a weekly digest email, and a separate marketing site with a blog and a science page that labels each mechanism as published theory or the app's own modelling choice.
- An MCP server
AI assistants can list activities, log them, read today's scores and delete a log through the app's own MCP server.
cold-start probe: 12/12 requests answered, slowest 1.58 s - The free Dopamine Audit
12 taps, about 60 seconds, no signup: one number for how your reward system is balanced and the lever that moves it most. Each result gets its own share page and image, and saving it leads to a 14-day plan.
- A model with memory that learns from you
A simulated two-year life showed the old engine forgot each day overnight. The rebuilt engine carries the last weeks forward, weighs dose and sleep debt, and grades the evidence behind each activity. Check-ins then tune the score to how you felt, with a forecast for tomorrow, risky time windows and 14-day personal experiments.
39/42 simulation checks · 122 web tests pass - Native iPhone and Android apps
A SwiftUI iPhone app that reads Apple Health, with an Apple Watch companion and home-screen widgets, and a Kotlin Android app that can capture screen-time minutes and Health Connect data on opt-in. Both run the same engine as the web and are in private testing.
engine parity 21/21 on iPhone, 16/16 on Android - Privacy and security pass
Account deletion, full data export and consent withdrawal on web, iPhone and Android, and per-user quotas on the AI features.
- Longer, sourced articles
Thirteen short blog and guide pages rewritten to about 2,000 words each, with cited studies and FAQ markup.
first 9 pages: 660 → 2,150 words on average
Who did what
The founder decided
- One number and one lever. No XP, no avatars.
- The app never tells you to want less: effortful rewards count as protective, and the science page labels that a modelling choice.
- The audit result is a self-reported index, not a medical test. Myo models dopaminergic state from behavior; it does not measure dopamine.
- Recovery data is processed only with explicit consent, and you can revoke it at any time.
- One calm, smart voice on the site, in the app and on share cards.
AI built and checked
- The web app and the marketing site, built in Lovable from the founder's briefs.
- Claude Code: the engine rework and its two-year simulation, the iPhone, Apple Watch and Android apps, the parity checks between them, and the privacy pass.
- The coach is an AI model grounded in your own scores and your last 7 days of logs.
- Long-form articles researched with AI, with the studies they cite checked before publishing.
Stack
Want something like this for your team?
From one call to something live. The same method is in the Claude Operator Pack.
From the founder
You might also like

Shares React · Edge functions

Shares TanStack Start · SwiftUI

Shares React · TanStack Start