Twenty questions, answered with how sure you are, produce the one chart almost nobody has ever seen about themselves: your personal calibration curve.
20 statements. For each one, answer true or false and say how confident you are. At the end you get your personal calibration curve — the gap between how sure you felt and how right you were. Almost nobody ever gets to see that about themselves.
No account. Nothing leaves your browser unless you choose to save it.
Calibration is the match between how confident you say you are and how often you're actually right. Someone who says "90% sure" and is right 90% of the time is perfectly calibrated. Most people aren't — and almost nobody gets to see their own gap.
The mean squared gap between your stated confidence and the outcome. 0 is perfect; always saying 50% scores 0.25. It rewards being right AND knowing how sure to be.
Every time you accept an AI's output, you're making a confidence call. Research on forecasting (Tetlock's work) suggests calibration is one of the most trainable, transferable judgment skills — and it's measurable from real work, which is exactly what Miatz Tuukul does.
This lab uses general-knowledge questions. Miatz Tuukul measures the same curve from your real AI work
Miatz Tuukul elicits a prediction before your consequential AI runs and reconciles it against what actually happened — a calibration curve computed from your real work.