Why ddd-studio?

ddd-studio makes explicit what actually happens inside a piece of software. Not as an end in itself, but because that's exactly where projects fail today: at the missing shared understanding. And not just between people — increasingly between people and machines, too.

The problem: prompts aren't a specification

Vibe coding has a blind spot. What a piece of software is supposed to do is often described far too imprecisely in a prompt — a few sentences, a rough idea, a lot of context that only exists in the author's head. An LLM fills the gaps anyway. It rarely asks for clarification; instead, it makes a plausible assumption and produces code.

The result runs, looks correct at first glance — and still diverges from what was actually meant. The expectations of the person prompting and what actually got built drift apart, without anyone noticing. That's the insidious part: it often goes unnoticed, because both sides believe they understood each other.

Shared understanding as the foundation

This is exactly where ddd-studio comes in. It forces domain concepts to be made explicit before any code is written: What Aggregates exist? Which Commands trigger which Events? Who is an Actor, what is a Policy, what is purely a read model?

This model isn't a diagram for decoration. It's the place where people — domain experts and developers — and machines agree on the same description of the domain. Once the model exists, there's no more interpretation gap that could be silently filled in.

A model is a very precise prompt

That makes a model built with ddd-studio something that combines remarkably well with today's AI tools: it can be understood as a very precise prompt. Instead of describing in prose what a piece of software should do, the model specifies Aggregates, Commands, Events, and their relationships exactly — machine-readable and with no room for interpretation.

In practice, that means the JSON export or the .esdm.yaml files can go straight into the repository. They're documentation and specification at once — versionable, diffable, part of the code instead of a separate wiki page that goes stale. An LLM working with this model in context no longer has to guess what was meant. It can look it up.

The real payoff

ddd-studio doesn't promise that software writes itself. It promises something more modest and, at the same time, more important: that everyone involved — people and machines alike — starts from the same explicit description of the domain. Everything else follows from that almost on its own.