AI hints
AI hints that teach without giving the answer
A learner stuck on an error or a failed test can ask for a hint: an explanation of what went wrong, or a nudge towards the fix, never the solution. Hints are off until you turn them on, never appear in exams, and your staff can read every one.
Last checked
What can learners ask for?
- Explain this error: after a code run in a lesson fails, an explanation of the error in plain words, from the code they actually ran.
- Three levels of hint on a wrong answer: "Get a hint", then "Another hint", then "Show me where". The first two never contain code. The last may point at the part of the code to look at, in at most five lines.
The AI sees the question, the learner's code, the visible tests with what their code printed, and only how many hidden tests failed. It never sees hidden tests, the learner's name or their email.
What controls do staff have?
Off until you turn it on
An admin turns AI hints on for the organisation in Settings, and can turn them off for any batch.
Never in exams
Exams never offer hints. Practice quizzes offer them unless you turn them off; graded quizzes only if you allow it.
Every hint visible
Staff see each hint beside the learner's code replay, and a page lists every hint with exactly what the AI was shown.
No chat box
Learners press a button; they cannot type to the AI. There is nothing to talk it into, and no conversation to drift.
Checked before it is shown
Every reply is checked for code before a learner sees it. One that breaks the rules is asked for again, then shown without its code.
A safeguarding route
If a learner's code suggests they may be at risk, they get support information instead of a hint, and your safeguarding lead is emailed a link.
Why hints rather than an AI tutor?
An open chat makes it easy for a learner to ask for the answer, and hard for an instructor to know what help was given. Fixed hints at three levels keep the learner doing the work, and the record of every hint lets an instructor see who is stuck and where, before the next class.
For assessment, that record matters too: an employer client can be told which answers were written with hints and which without, because hints never appear in exams.
Questions and answers
Will the AI give learners the answer?
No. It is told never to write the answer, the first two levels contain no code at all, the third may point at one block of at most five lines, and every reply is checked before it is shown.
Which AI model writes the hints?
Claude, made by Anthropic, used through its API with processing in the United States. The model gets the question, the code and its output, never the learner’s name or email.
Can instructors read the hints?
Yes. Admins see every hint, instructors see hints in their own batches, and each hint shows what the model was given: the question, the code, the visible tests and how many hidden tests failed.
How many hints does a learner get?
Up to 100 a month on a subscription and 30 on the trial and free plan, with at most 10 at a time and one more every few minutes, so a learner cannot ask for hint after hint without thinking.
Can learners under 18 use AI hints?
From 16, if the training provider turns them on. Hints are labelled as AI and come only from fixed buttons, and the safeguarding route applies.
Do AI hints cost extra?
No. They are in every price band and the free trial, with no per-learner AI add-on.
Try it with your next cohort
Start a 30-day free trial with up to 10 learners and no card, or tell us what you teach and we will show you The Skill Site with your own exercise.