Case study · Web + iPhone · My own product
Every free program online assumed a gym, five days and someone else’s goal. I wanted a plan built from my actual answers.
Decisions that mattered · 01
- A rules engine, not a chatbot. Same answers, same program, every time — and I can explain every choice the engine makes. AI assistance is on the roadmap as an adjustment layer, not the foundation.
- Calendar export as the finishing move. A program that lives in a web page gets forgotten. An .ics file puts each session in the user’s week.
- Test-first. The engine and the iCal generator have pure-logic test suites; the API has integration tests that provision a throwaway database and run against a real PHP server.
- No build step. Vanilla ES modules and hand-written CSS. Deploys are file copies.
Stack
Vanilla HTML/CSS/JS front end, PHP 8 + MySQL API, node:test for logic, shell-driven integration tests. cPanel Git deploy.
Outcome
Live on the web; iPhone app submitted September 2026.
Screens · 02
