A 30-question reasoning test (matrices, number series, mental rotation, odd-one-out) with questions generated on the spot, so nobody can memorize answers. Your score is free and comes with its margin of error, scored with item response theory. No subscription, no locked results.

Hi Product Hunt! I'm Airton, the maker of Lume IQ.

I built it after looking at how most online IQ tests work: you answer the questions, then your score is locked behind a cheap "trial" that quietly turns into a monthly subscription. I wanted the opposite.

How Lume IQ works:

- 30 questions across 4 domains (matrices, number series, mental rotation, odd-one-out), generated on the spot from rules, so every session is different

- Scored with item response theory: you get an estimate plus a 90% margin of error, not a single magic number

- The score is free. There's an optional one-time report ($9.90) that explains every question. No subscription, 7-day refund

- Available in English and Portuguese

To be clear: it's an estimate for self-discovery, not a clinical assessment, and our norms are still provisional while we collect more data.

I'd love your feedback on two things: did any question feel unfair or ambiguous, and does the result page make the margin of error clear?

A question I'm getting a lot: how accurate is it compared to traditional IQ tests?

Honest answer: Lume IQ is not a clinical test and hasn't been validated against WAIS or Stanford-Binet, so think of it as a reasoning estimate, not an official IQ.

It uses 30 items across 4 areas, scored with item response theory and placed on the usual IQ scale (mean 100, SD 15). Norms are still provisional and we keep calibrating them as more people take it. That's why every result comes with a 90% interval (usually around 8 points either way) instead of a single number, and why fast or careless sessions are flagged as inconclusive instead of scored.

Full methodology: getlumeiq.com/metodologia