we built an AI that judges peoples AI skills. heres what keeps surprising us

building aisa.to (conversational AI skills assessment) and still finding patterns that catch me off guard.

the latest one: people who use AI 30+ times a day consistently score lower on output verification than people who use it 5 times a day. the heavy users develop a kind of automation blindness — they stop checking because theyve "always" been fine.

meanwhile the 5x/day crowd tends to treat each interaction more deliberately. they verify more, iterate more, and catch errors the power users miss.

were not even measuring how much someone uses AI. were measuring how they think about using it. turns out those are very different things.

still early but this keeps showing up. curious if other founders building AI tools are seeing similar patterns — the users who look most "advanced" on paper sometimes arent.

building at https://aisa.to if you want to see what your own AI skill profile looks like.