There is an easy version of an AI tutor: type a question, receive a complete answer, copy it down. It feels productive and it teaches almost nothing. The moment a tool hands over the finished solution, the student's thinking stops. We built the Socratic Tutor to keep that thinking going.
Questions before answers
In Learn mode, the tutor's first response to a hard problem is rarely the solution. It is a question: What do you already know here? Which formula might apply? What happens if you try the first step? The aim is to find the exact place a student is stuck and give just enough of a nudge to get them moving again — not to do the problem for them.
Hints that fade
Hints are graded. A student who is nearly there gets a small prompt; a student who is genuinely lost gets more scaffolding, and the tutor will work through a similar example before returning to the original problem. If a student insists on the full answer, the tutor will eventually give it — but it then asks them to explain it back, because being able to reproduce a method is the point.
Checking for real understanding
After a concept clicks, the tutor follows up with a quick variation to confirm the understanding is durable and not just pattern-matched to one example. This is the part students skip on their own, and it is where most revision quietly fails.
Why this matters for exams
Exams reward students who can apply a method to an unfamiliar problem, not students who can recall one worked example. A tutor that guides instead of solving builds exactly that transferable skill. It is slower in the moment and far more effective over a term — which is the trade we think every school should want.