How Webforgia AI Works
Webforgia's AI writes data, not code — and that one decision is why its output is consistent, checkable and safe to publish.
Data, not code
For websites and the other document-like types, the AI never writes the page’s HTML. It writes a Site Specification: which sections, in which order, with which words, images and theme. That document is validated against a strict schema, and a fixed, tested renderer turns it into the page.
Why that matters to you
- It cannot break the layout. A section is a designed block; the AI fills it, it does not invent markup.
- Edits are precise. Changing a specification is a targeted change to one field, so an edit can keep everything else exactly as it was.
- It is safe to publish. Published sites run no AI-written code, and every link is checked before it can be stored.
The same idea, everywhere it fits
Cross-platform apps and 3D games follow the same pattern with their own specifications — an App Spec turned into a React Native project, a Game Spec driving a Three.js runtime. Only the sandbox builds (instant web apps, 2D games, animations and spreadsheets) have the AI write a program, and those run in a locked frame with no access to your account or the network.
Code and AI, side by side
Most of a build is not AI at all. Each stage is either plain code (the same answer every time, and free), an AI call, image generation, or a real browser looking at the page. A website build, for example:
| Stage kind | Website stages |
|---|---|
| Plain code | 11 |
| AI calls | 3 |
| Image generation | 1 |
| Browser | 2 |
Every stage, for every build type, is in the Generation Pipeline.