Nikita Timofeev

I've spent years writing the same CRUD boilerplate for every new backend, and I keep noticing the same pattern in the tools meant to fix it: scaffolders generate a project and abandon you the moment the schema changes, and low-code removes the boilerplate but locks the product into someone else's runtime.
I've been working on a third way: a language-agnostic code generator. You describe the domain once in a small YAML model, and it produces real, readable source code you own MIT licensed and open source. The core is deliberately language-neutral: it parses and validates the model, builds an intermediate representation, and renders it through pluggable bridge templates. The first bridge we ship generates ASP.NET Core + EF Core + PostgreSQL but any other language is just a new bridge (templates and type mappings), with zero core changes.
The part I care most about is "day two". Business logic lives in separate files that survive regeneration, and the migration engine renames database columns instead of dropping them, so a field rename doesn't destroy data.
Before I go further, I'd genuinely like to learn from people who have been through this. What made you give up on a code generator or low-code platform day-two maintenance, lock-in, or the quality of the generated code? And if "any language, you just write a bridge" is the model what language would you want first?
Happy to dig into how it works under the hood the IR contract, deterministic rendering, schema snapshots, and a compliance suite (TCK) that certifies third-party bridges. Product and launch details here for context:

https://www.producthunt.com/prod...