How layers fill in — mock vs. generated
Kiln's model is a stack of layers — capabilities, areas, entities, behaviour, and so on — each derived from the ones above it. A common misunderstanding is that describing your business "generates everything" in one pass. It doesn't, and knowing the real mechanism makes the whole tool easier to read.
Describing your business does not generate the model
When you write or import your description, Kiln runs an understand pass: it mirrors your text back as a short summary and a list of open questions. That's all. No capabilities, entities, or any other layer are produced at this point. It's a comprehension step, not a generation step.
Every layer shows a live draft before you generate it
So why does every stage already show something when you open it? Because each layer that you haven't generated yet displays a live, automatic derivation of the layer above — a deterministic best-guess computed on the spot, with no AI involved. It exists so stages are never blank and you can see the shape of things, but it is placeholder scaffolding, not model output.
The real content for a layer is produced only when you open that stage and click Generate with LLM (or, on some layers, Enrich → Web). That runs the AI, and its result replaces the draft and is saved.
This gives every layer three states, shown as a dot on the stage rail:
| Dot | State | Meaning |
|---|---|---|
| ○ | Empty | Nothing here yet. |
| ◐ | Draft | A live, auto-derived placeholder. Not generated — click Generate to make it real. |
| ● | Generated | The AI produced this layer (or you authored it). This is real content. |
The shape tells you how far along a layer is; it is never a warning. A draft (◐) is simply a layer waiting for you — Kiln shows no findings or problems on a draft, because there is nothing to judge until the model has actually produced the layer.
Health is shown only on generated layers
Once a layer is generated (●), Kiln checks it and shows its health as a small coloured badge next to the dot:
- no badge — clean; a green ● means "generated and looks good";
- amber badge — minor suggestions worth a look;
- red badge — concerns you should resolve.
The number on the badge is how many open findings there are. Because health appears only on generated layers, a coloured badge always means the AI produced something and there's a real point to act on — never a false alarm about a placeholder.
Work top to bottom — regenerating resets what's below
Layers depend on the ones above them, so build in order, from the top. One consequence matters: regenerating a layer resets the layers beneath it back to drafts. For example, regenerating Capabilities clears the generated Entities, Behaviour, Roles, and everything downstream — they return to ◐ drafts until you generate them again. Kiln warns you before a regeneration would discard authored work below (see Protecting your fixes). Generating forward — the next layer down — is always safe.