Tezvyn

Everyone who has interviewed for a tech role has had the same forty seconds. You know the answer. You've shipped this exact thing. Then you're asked to say it out loud, and what comes out is vague and circular — and you watch the interviewer's attention leave the room.

The gap isn't knowledge. It's retrieval, under pressure, out loud, to someone who is judging you.

Tezvyn is built for that gap. One idea per card, sixty seconds to read. Interview cards name what the question is really testing, the answer that lands, and the red flag that ends it.

Built by one person, self-funded. 8,600+ cards, free to read on the web.

**Mobile** — Flutter 3 / Dart 3, Android. Riverpod, go_router, freezed, Material 3, sqflite for offline reading.

**Backend** — FastAPI 0.115 on Python 3.12, SQLAlchemy 2.0 async over asyncpg, Postgres 16, Alembic, Pydantic v2, Sentry.

**Web** — Next.js 15 App Router, React 19, TypeScript, Tailwind. ~8,700 pages built on demand via ISR.

**Content** — cards are LLM-generated against a fixed template, then machine-validated and deduplicated before publication. Disclosed on the site.

**Infra** — Docker Compose on one Hetzner box, Caddy 2 in front. Deploys are a manual SSH command. No Kubernetes, no queue. One box is enough.