Quest Ladder
TeachFORGE.ai
Quest Ladder / Preview and verification
Preview, new examples and render verification
See what a learner will actually get, several times, before anyone else does.
A type you have not sampled is a type you have not written.
Preview and refinement
The Review stage opens on Preview and AI refinement: a real generated instance of your type, rendered as a learner would see it.
- Refresh preview re-renders the current sample after an edit.
- New examples samples fresh values, so you see a different version. It is labelled Example 1, Example 2 and so on.
- A toggle switches between the tokenised template and the resolved preview that matches the student view.
Press New examples at least five times before you publish anything parametric. Bad combinations hide: the division that comes out whole four times and awkward on the fifth, the diagram that only overlaps when two variables are both large. One sample proves nothing.
Render verification
Render verification checks that every part of the type actually draws: the LaTeX compiles, the diagram renders, the chart has its data, no block fails. The badge reports one of four states.
| Badge | Means |
|---|---|
| Verified — clean | Everything rendered. |
| Needs re-verification | It was clean, then you edited something. Run it again. |
| {{count}} render errors | Something does not draw. The list says what. |
| Not yet verified | It has never been checked. |
This is a gate, not a warning. Publishing is refused while errors remain, and the control says why: fix the render verification errors before publishing.
What the errors usually are
Three causes account for most of them.
- A LaTeX or TikZ expression with a syntax error, often from a stray brace after an edit.
- A token pointing at a variable that has since been renamed or deleted, so nothing resolves.
- A diagram whose coordinates come from variables that can produce a degenerate case — a triangle with no area, a chart with no range.
The third is the one to watch, because it passes on most samples. A verification run that samples several sets of values is what finds it, and it is exactly the failure a learner would meet once in twenty.
Mark ready
Mark ready sets the status to published once you are happy with the sampled preview. It is the moment the type becomes listed and assignable — see Publish, unpublish and the completeness chips for what that changes.