In build · 2026-04 → ongoing
Social Sine
Shipping a real product solo with AI augmentation — and the discipline that takes.
The bet
A real, multi-platform social product, built and shipped by one person using AI as a force multiplier. Not a demo. Not a prototype. A working app with a real database, real auth, real scope. The case study for the methodology — AI-assisted software, done with discipline.
What it is
A social app where the product is the meet-up, not the scroll. People post public or scoped meet-up intents; the app routes intent to the right people and gets out of the way. No follower counts, no like counts, no engagement metrics surfaced.
Decisions that mattered
- Scope + preference split. Scope says who can see; preferences say what you want to see. Two orthogonal axes — never collapse them.
- Region as a primitive. A named geographic area with a radius, independent of the user’s km slider.
- Symmetric typed connections. No follower graph. Only the type of relationship.
- DB-enforced admin guard. Role changes require a session-set variable + trigger.
- AI as a teammate, not a replacement. Every commit reviewed before it lands. The AI generates; I decide.
What’s next
Phase 2 is the offline moment itself: arrival, identification, the “it happened” signal. The app should go quiet during the meetup window — that’s the test of whether the design filter held.