Book a demo
Book a demo
Home›Reasoning
Most integrity tools try to catch cheating after it has happened. We built a reasoning test where cheating does not pay, so candidates stop trying. The question is only visible while the candidate is physically holding it on the screen. Let go, to reach for a phone, open another window or take a screenshot, and it disappears while the clock keeps counting; a screenshot captures an empty screen. Working the answer out is faster than outsourcing it, and the way a person moves tells us when someone has not. Verbal, numerical and abstract reasoning, adaptive, and built for unproctored use in the age of language models.
◆ 01 · Why it works
The industry's instinct has been to look for a kind of question a language model cannot answer, and build a product on it. That position keeps collapsing; every benchmark designed to be hard is solved within months. So we built a test where the difficulty of outsourcing lies in the interaction, not in the item.
Underneath that mechanic is an ordinary adaptive test with a calibrated item bank. A practice pair sets the starting point, the difficulty ladder settles where the test learns the most about the candidate, and every item is scored on its own difficulty, so the route someone took does not change what the score means. New items are calibrated against the bank quickly rather than after a season of trials, and the bank is refreshed on a regular cycle, so no item stays in circulation long enough to leak.
Item on screen. Clock counting.
Screen empty. Clock still counting.
The clock never speeds up or pauses. Letting go costs you the question, not the time.
Every reach is recorded: how straight the path is, the small corrections along the way, where the peak speed falls, the natural tremor of a human hand, and what was hovered over before the answer. People have a signature. Scripts do not.
The orb feels slightly different every session. A person adapts over the first few reaches. A script replaying a learned movement either does not adapt or adapts wrongly.
Interaction authenticity returns a score with reason codes for a human to review. No adverse decision rests on a movement classifier alone.
◆ Next step
The fastest way to see what this looks like is to give us one job. Send a job description and we'll show you what the first scenario looks like for it.