Sounds right. Is it?
Pick something you know cold. An AI writes eight confident claims about it. Half are true. Half it made up on the spot, in the same voice. You get ten seconds a claim. Then a second AI, which never saw the answer key, tells you how sure it was of each one.
Nothing you type is stored. The topic goes to the AI, the claims come back, and that’s the whole transaction.
Why this exists
The fastest way to learn what AI gets confidently wrong is to test it on something you already know. On a topic you’re an expert in, you can feel the difference between a claim that sounds right and one that is right. On a topic you’re not, you can’t. That gap is the whole skill.
We built this as a demonstration for myQuest, a skill-building platform where exercises like this run inside a learning mission. The version here plays on national parks and whatever you type in. The same mechanic runs on a company’s own material: its policies, its product line, its procedures, with the claims written from that material instead of from general knowledge. It sits inside a mission next to the practice and skill mapping the platform already does.
It is one exercise type. Timed judgment calls, resource trade-offs, branching decisions, and head-to-head rounds all fit in the same frame. If you have training content and want it to prove something, that’s a conversation we like.