Quest Ladder
TeachFORGE.ai
Quest Ladder / Generating questions
Generating questions from a type, and batches
Produce as many concrete questions as you need, and see them before they are saved.
Generate many. Read some. Save the good ones.
Type, instance, batch
A question type is a design; a question is one instance of it with the changing values filled in. Generate questions from a type is where instances come from.
The steps are: select the type, choose How many questions?, click Generate prompt, then copy or run it. You can add notes for the batch — the placeholder suggests things like avoid zero, use brackets for negatives — which is the quickest way to steer one run without editing the type.
Factory types
A fully parametric type behaves differently and the page says so: it is a factory, samples generated from it publish immediately, and live teaching mints a fresh sample each time anyway.
For those you rarely need to generate a bank at all. Generate a handful as samples to check the parameter ranges, and let the delivery produce the rest.
Preview before saving
Whichever route produced the JSON, it comes back into Import bulk JSON and Parse & preview shows a table — Stem, Answer, Hints — with View on each row to see the full preview or the raw JSON.
Review them here, remove any you do not want, and then save. The page offers saving all as drafts or saving and publishing all.
Read at least a few before saving, and read the answers rather than the stems. A parameter rule you did not write shows up as an answer that is negative, fractional or absurd, and it is far cheaper to catch here than after a class has met it.
If some rows had errors the import says so and imports the rest, with warnings. Those rows are usually a malformed reply rather than a bad question, and re-running them is quicker than repairing the JSON.
Batches
Batches lists generation runs with when they were created, what kind they were, their status and their progress. A run can be opened to see what it produced, cancelled while it is going, or deleted afterwards.
Run in browser is how a batch executes against your own GPU: your browser does the talking, so the tab has to stay open. Platform batches can finish after you leave.
This is the page to check when a questions count jumps unexpectedly — it is almost always a batch that ran twice.