is the gap between what you claim to know and what you actually know. Perfectly calibrated thinkers who say 80% are right 80% of the time. Everyone else — even experts in their own domain — systematically overstates certainty.
Lab makes the gap visible. Every challenge asks for both an answer and a confidence rating. Over time, a accumulates, and you can see your curve — the line of 'stated confidence → actual accuracy' drawn from your own data. The curve is the feedback. Nothing else is as direct.
This is the mode the Good Judgment Project used to identify superforecasters — the tiny slice of forecasters who outperformed intelligence analysts with classified data. The difference between them and everyone else wasn't raw accuracy. It was . They knew when to trust themselves. Most people don't.
This trains
The match between stated confidence and realized accuracy, measured by . A tight match means your inner voice is a useful instrument.
How a session feels
4 steps, 3–5 minutes. Repeat until the feedback starts shaping your instincts.
- 1
Answer a question
Mixed domains — trivia, forecasting, reasoning. The domain matters less than the confidence rating.
- 2
Rate confidence 0–100%
This is the measurement. Lie to yourself here and the whole mode stops working.
- 3
See the Brier delta
A lower score is better; 0 is perfect. The delta this session vs. last session is what you watch.
- 4
Read the calibration curve
After 30+ rated attempts, your personal curve emerges. This is the artifact you're training.
Example challenge
A representative prompt. The real session varies the difficulty and format.
Which city is farther north — Chicago or Rome?
- Chicago
- Rome
Why this answer
Rome (41.9°N) is farther south than Chicago (41.88°N) — barely. This is a classic question: 'obvious' answers flip. Rate accordingly.Who it's for
- Forecasters, traders, and VCs — your livelihood is a distribution, not a point estimate
- Leaders making high-stakes calls under uncertainty
- Anyone who has ever said 'I'm absolutely sure' and been wrong — you are the ideal user
Try a challenge — no sign-up
The demo pulls from the public challenge bank. Your confidence rating and result are the same mechanics you'll see in the real mode.
Start with Calibration Lab
Your first session generates a score baseline in under 10 minutes.