Instead of building infrastructure (connectors, vector DBs, transcription, entity extraction, etc.), I reused Graphlit, the same stack our customers use. That saved about 4-5 months of backend work and let me focus 100 percent on product and UX.
The result: I built and shipped an MVP in 8 weeks, got the first paying customer in 4 months, used 30+ data connectors with real-time sync and entity-aware semantic search, and 100 percent of coding was assisted by AI (Claude Code and Factory Droid).
The post covers the full stack (Next.js, TypeScript, Graphlit SDK, Vercel), the workflow of building with AI coding tools, what worked, what didn't, and why platform founders should build on their own platform.
TL;DR: If you've built a platform, use it to build products. It's the fastest way to find product-market fit, and AI coding tools make it shockingly fast.