The first project I built almost entirely with AI

I'd been watching Cursor and AI coding models get better for months, but I hadn't really tested the limits. Toolable was my excuse to find out: how much of a real, production platform could I ship if I let AI handle most of the actual code?

The platform is a no-code marketplace for AI content generators. You describe the tool you want — a blog post outliner, a product description writer, whatever — and AI creates a fully functional generator with custom input fields, optimized prompts, and a shareable URL. Other users can discover it, use it, fork it, and build on top of it. Think GitHub, but for AI tools.

There's a meta quality I like about the whole thing: the platform uses Claude to build the generators, so it's AI creating AI tools. When a user fills out two fields describing what they want, Claude generates the complete specification — system prompts, input configurations, categorization tags, everything. The tool that creates the tools was itself mostly created by AI.

I still owned the architecture. I set it up as a Turbo monorepo with separate packages for the database layer, the generator logic, and the web app. I went with Drizzle ORM over Prisma for the PostgreSQL setup because it gave me tighter TypeScript integration and more control over query building. I added trigram indexes for fuzzy search so the marketplace could handle discovery without pulling in an external service like Algolia. Every meaningful technical decision was mine. But the implementation — the actual lines of code — was mostly Sonnet 3.7, with me directing.

The original thesis had a programmatic SEO angle: every generator a user creates becomes an indexable page with a unique URL, so the site's search footprint grows organically as the community builds. I'd seen other AI tool directories ranking well on Google and figured a social layer on top of that model could compound nicely. I launched it on Product Hunt and it's live at toolable.ai.

What I didn't expect was how much this project would change the way I think about building software. Before Toolable, AI coding assistants felt like a nice autocomplete. After it, they felt like a genuine force multiplier — something that was going to reshape how everything gets built. I went from curious to convinced over the course of a single project, and pretty much everything I've built since has been shaped by that realization.