Custom where it matters. Standard everywhere else.
Why our platform makes the specification the product, not the code.
Problem 1
AI writes code, not reliable software
of AI changes that silently altered behavior were approved by the same model in self-review.
Reddy et al., 2026AI types faster. Someone still has to understand, verify and maintain it.
Problem 2
Low-code, the gilded cage
A fast start. And then:
The platform decides what your app can do.
Customizability and testability end where the platform ends.
Leaving usually means rebuilding.
Turning business users into developers rarely works out.
The real problem
Two kinds of change
Processes, requirements, organization.
Frameworks, libraries, UX standards.
Both hit the same codebase.
How the platform works
Specification first
Your application is described in BDL (Business Definition Language): a structured, open model of data, processes, roles and rules.
Readable
Business teams, developers and AI agents work on the same model.
Verifiable
The model can be checked automatically. Gaps and contradictions show up before any code exists.
Generated, not improvised
AI helps build the model. A deterministic generator turns it into code in seconds. Same model, same result, without using a single token.
Decouple business logic from technology
Your team only owns its custom code.
Change the model → regenerate = new feature
We update the platform → regenerate = same stack, up to date
New framework and library versions, security fixes, UX improvements.
Automation without dead ends
Every step adds control, and responsibility.
Work on the model: no code to own
- 1
Pick features
Choose what you need from the standard model.
- 2
Change the model
Adapt and extend it with your own model changes.
Write code: you own it
- 3
Extension points
Plug in code with fixed signatures: actions on triggers, filters, event and error handlers, workflow scripts.
- 4
Custom components
Replace standard UI components, such as layouts or dashboard widgets.
- 5
Custom modules
Add your own sections with hand-written backend and frontend, integrated into the app.
- 6
Take over generated files
Switch off generation for a single file and maintain it by hand. Keeping it in line with generator updates is then up to you.
As much standard as possible, as much control as needed.
One model, two kinds of users
An IntelliJ plugin, BDL models as readable files in Git, your own coding agent, and a generator for database, backend and UI.
Specs Architect in the browser: capture requirements, change requests and bugs, manage labels and help texts, and test the current build.
The lever isn't less typing.
The lever is less code you have to own.
Let's model one domain of your landscape together.
We show you what comes out of it. Or call us: +49 175 210 7271