Vocalize.Cloud: build and deploy an AI SaaS in two days
The speaker outlines a repeatable solo-founder playbook demonstrated by his Vocalize.Cloud idea: validate before you code, pick a productivity-first tech stack, wire minimal backend automation, and monetize with a simple pricing model. He stresses the mindset — "make your own luck" — using Danny Postma's Headshot Pro (12,000 customers) and anecdotes (Quentin Tarantino, Dax Lam) to show opportunity arrives to those prepared. He will validate demand via his YouTube audience within 24 hours before investing further.
Technical choices are pragmatic: choose tools you like. He prefers 'spelt kit' and Firebase (the "f**k it stack"), uses the Carbon design system or tailwind for UI, and wrote 'spelt fire' to bind spelt stores to Firebase. Key architectural flow for Vocalize.Cloud: user submits form -> Firestore document written -> Firebase Cloud Function triggers -> call 11Labs API for voice cloning -> save audio to a cloud storage bucket -> generate a secure per-user download URL -> update Firestore status -> front end (Firebase SDK + spelt store) reflects real-time progress. Security is one-line Firebase rules restricting users to their own data.
Database and infra tradeoffs: Firestore fits simple user-to-audio relationships, but relational DBs (MySQL/Postgres) are safer for complex relations; cautionary tale: CodeDamn hit scale limits with Prisma + PlanetScale. For billing, he uses Stripe checkout sessions generated by an authenticated Cloud Function and a webhook Cloud Function that credits tokens in Firestore and sends confirmations via SendGrid. Product mechanics: 100 free tokens at signup; 11Labs charges 18 cents per 1000 characters, so he prices at 69 cents per 1000 characters (a Versailles strategy) to ensure margin.
Concrete names and numbers: $20 million a year (one-man Chrome extension business), Peter Levels earning at least $3 million a year, 11Labs API, Firestore, Firebase Cloud Functions, Stripe, SendGrid, and a promised spelt kit + Firebase course by end of June. The repeated prescriptions: validate early with real users, prioritize execution and marketing over shiny tech, and pick tools that maximize speed of iteration.