How it works
Hints, not answers.
Every problem runs the same three phases a real interviewer uses, then grades the conversation rather than the answer. You can’t skip to the code, because in an interview you can’t either.
Say it before you type it
You explain your plan in plain English — no code. The tutor asks one question at a time and gives one hint at a time, and reveals the core insight only after three honest tries. If your plan is wrong, you find out here, which is where it’s cheap.
- One hint at a time
- reveal after three honest tries
Defend the cost
You state time and space. When you’re off, the tutor doesn’t correct you — it asks how many times each element gets visited, until you can get there yourself.
- No correction, only questions
Then make it run
You write Python in the editor and run it against per-constraint hidden tests. The reference solution and the expected outputs never leave the server, so passing means your code works, not that you matched an answer key.
- Hidden tests stay server-side
- run as often as you like
The conversation is what’s graded
You get a difficulty-weighted verdict — solving a Hard with one hint outscores a clean Easy. It’s computed on the server from your recorded hints and harness pass rates, not from the tutor’s opinion of how the session felt.
- Difficulty-weighted
- recomputed server-side
What it costs
One problem, start to end
Credits come out once, when the problem appears — then the whole conversation is paid for. Step through it and watch the ledger.
Step 1 of 7
Your month starts
Your plan drops 30 credits in. They are yours to spend on any mix of problems — nothing is earmarked by topic or difficulty.
Your balance
You can’t go negative: if a problem costs more than you have left, it’s refused before it starts rather than overdrawn. See pricing for what each difficulty costs.