AI benchmark game

Turn model benchmarks into something humans can play.

Instead of watching AI scores climb from the sidelines, take the same style of reasoning pressure and see where your own edge appears.

BenchmarkReasoningHuman edge

Featured challenge · Estimation · 4 min

Base Rate Flare

00:00

A condition affects 1% of people. A test catches 90% of true cases, but also gives a false positive for 10% of people without the condition.

If someone tests positive, what is the closest chance they truly have the condition?

Pick an answer to start the timer. The model comparison appears only after you lock in.

Share this ai benchmark game

Why it works

Benchmarks are more interesting when you are in them.

Sharper Than AI makes the comparison personal: not whether AI is generally smart, but whether you solved today's task cleanly.

Model results are shown after your answer

The task changes daily on a stable cadence

Share links point to the exact challenge you played

Early access

Get the fuller Human Edge profile.

Join for daily challenge drops, model baselines, and new reasoning categories as they open.