I created a product compiler with DSLs

Hello HN! I've been working on TypMo as a solo builder for past 5 months or so. It started as a wireframe-to-prompt tool but has evolved to what I call as a "product compiler". TypMo takes whatever you have as product input (a prompt, customer interviews, an existing codebase) and turns it into a structured, queryable spec graph. It synthesises into Typed nodes (personas, business rules, agent blueprints, entity state machines, user journeys) with relationships you can traverse via MCP. Under the hood, synthesis is a compile pipeline through DSLs, each layer stricter than the previous: input parsing, typed graph construction (roles, rules, entities, journeys), agent orchestration, and the final spec graph. Validation isn't an LLM grading another LLM. It's a fixed set of checks: different ways the spec can contradict itself, a formula that scores whether everything wires together (every rule has an enforcer, every entity has a state machine), and a 14-pass structural validator. The bet is that intent should compile, not just be transcribed.

I'm getting some good feedback for the product but I wanted to understand the community's perspective. I believe a spec graph is better than static md docs. But want to hear more perspectives.