Learn the method with a similar worked example
Unrestricted chat often answers the homework. RubricMark’s Example button regenerates a parallel worked problem with a Pro model — independent of Rubi hints — so you still attempt the real item.
Recent evidence on AI tutors is blunt: unrestricted chat can raise assisted practice scores while hurting unassisted exam performance. Hint-only tutors fare better. RubricMark’s learning posture follows that line — AI enhances traditional tools; it does not replace retrieval.
The mid-sit Example control sits in the test runner next to Rubi, but it is a different product:
- Rubi — flash-lite Socratic nudges; never the keyed answer.
- Example — Pro model invents a similar question plus a full worked solution; your sit item stays unspoiled.
What the model is instructed to do
- Read the skill / method template from the bank item (server-side).
- Change numbers, names, and surface details.
- Return a new stem, solution steps, and final answer for that twin only.
- Reject copies of the reference stem.
Credits and caps (so it stays a study tool)
Each generate costs prepaid credits (list price on the drawer). Results cache on the attempt so reopening is free. Regenerations are capped per question and per paper — enough for method study, not infinite answer shopping.
A responsible use script
- Try the sit item cold.
- Generate one Example if the method is unclear.
- Study the steps; close the drawer.
- Solve your item without scrolling back to the twin’s final line.
- Use Rubi only if you need a Socratic nudge mid-working.
Bottom line
AI earns a place in the runner when it strengthens a traditional worked-example card — not when it finishes the homework. Example + Daily Practice Packs keep the attempt yours.