Live demo · runs in your browser

Try the evaluator

Answer the question below. Rubric scores it against an explicit rubric, cites the evidence for every score, and abstains rather than guess when the signal isn't there. This is a simplified taste of the real engine — no data leaves your browser.

Question · algorithms · difficulty 4/5
You have N people, and knows(a, b) tells you whether person a knows person b. A "celebrity" is known by everyone but knows no one. Find the celebrity in as few calls as possible — and explain why your approach works.
Scored on: correctness (50%) · communication (25%) · depth (25%)
Load an example: strong · partial · fluent-but-wrong · too thin
What you're seeing: the score is anchored to explicit criteria, not a vibe. Every rating quotes the text it's based on — no evidence, no score. A fluent, confident, wrong answer still scores low on correctness (the dimensions are judged independently). And a too-thin answer makes the evaluator abstain instead of inventing a number. That's what "defensible" means. ← back to Rubric