I built a programming language mostly with AI — and it somehow became a real full-stack too

Borisalv Marinov — Sun, 20 Sep 2026 19:35:39 +0000

I began BMEC mainly as a side project.

The aim was to find out whether it was possible to create a programming language which was easier to read, simpler for AI to handle, and at the same time useful for real-world applications.

I never really expected it to make much progress.

Now BMEC has its own .bmec files, compiler, type system, CLI, formatter, VS Code support, SQLite and PostgreSQL, APIs, authentication, UI, styling, and a number of actual test applications.

The fundamental concept is that BMEC should appear more like a set of clear instructions than ordinary code, even though it still has a strict compiler as a basis.

For example:

let age be 20

if age is at least 18
show "Allowed"
otherwise
show "Denied"

Under the hood, the structure is still normal compiler stuff:

.bmec source
→ parser
→ type checking
→ typed IR
→ runtime/backend

It is therefore not merely the case that "AI understands English and guesses what you mean."

The language is to remain deterministic.

I’ve also been trying to make BMEC full-stack, so one language can cover things like:

backend logic
APIs
database work
authentication
UI
styling
forms
browser events

I recently used it to create things such as a task dashboard and a job-booking app which included login facilities, different roles, PostgreSQL, browser end-to-end tests, and persistence.

In some areas it remains crude.

The UI/style system in particular requires a great deal of improvement, and BMEC is not at all close to the level of maturity or the range of available tools that Python, TypeScript, Go, and Rust have.

Well, that's actually the reason I'm posting this.

At first it was just a side project, but now I'm starting to question if it might really end up being useful should I continue with it.

I'd like to know what other developers have to say.

Does this appear to be an interesting course of action, or does it rather seem to be one of those projects which is enjoyable to build but probably of no use outside the experiment?

I’d especially like feedback from people interested in:

programming languages
compilers
web development
AI coding
UI/design systems
developer tools

I haven't yet worked out what BMEC ought to be, so at this stage any feedback could probably have a significant effect on the project.

DEV Community: Borisalv Marinov

I built a programming language mostly with AI — and it somehow became a real full-stack too