
The tutor that asks
Don't just get the answer. Understand it.
Socrat maps what you already understand, then asks the questions that build the rest, one layer at a time.
Most AI is built to hand you the answer.
You get the fact, fast, and none of the layers underneath it. Socrat asks instead, so you build them.
A typical AI
Fast. Correct. Forgotten.
Socrat
You work it out. It stays.

The first of three ideas
Understanding is built in layers.
Every idea rests on the ones beneath it. Socrat finds the lowest one you're missing and starts there.
- 4Long division
- 3Division
- 2Subtraction
- 1Addition
The second idea
An analogy borrows what you already know.
Something you know deeply gives a new idea somewhere to land. If you play tennis, slope might start like this:
“You're up 5–2, but she's won the last four points. Who's really winning?”
The third idea
Working it out is what makes it stick.
Stuck? Each step gets smaller. None of them is the answer: you reach that yourself.

- 1The question
- 2A nudge
- 3A smaller question
- 4A piece underneath
- 5A parallel example
- Your answer, found by you
Socrat maps your understanding.
Every answer adds to it: what you know and how deeply, what's missing underneath, what you love and where you get stuck. Not a score. A map that grows with you.

- Built twice on your own, a day apart
- Building getting there
- Not built yet where we start
- Known by name nothing underneath yet
Your map is yours to see, export or delete. How we treat it
The research agrees.
2×
the learning gains of a great class, from an AI tutor built never to give the answer
Harvard, 2025
−17%
on the exam after practicing with AI that just answers
PNAS, 2025
0.55
effect size of explaining ideas in your own words
Meta-analysis, 2018

