we're building an AI that assesses people's AI skills — biggest surprise so far
been building aisa.to for a while now. its an AI that measures how well people actually use AI tools through conversation instead of quizzes or self-reported surveys.
the thing that surprised me most: self-reported skill barely predicts actual skill. people who rate themselves 8/10 routinely show 4/10 practices when you actually watch how they work with AI. the confidence-competence gap is massive.
we score across 5 dimensions and 11 criteria, and the pattern we keep seeing is that output verification is the biggest differentiator. most ppl just accept whatever AI gives them without checking. the ones who are genuinely good have a system for evaluating what comes back, not just vibes-checking it.
other thing i didnt expect: the persona profiles are way more useful than raw scores. two people can score the same overall but have completely different shapes. a "Sceptic" who's cautious but underuses AI looks nothing like a "Copy-Paster" who uses it daily but never verifies anything. both need different growth paths.
still early but the data keeps getting more interesting. anyone else building in AI assessment or skills measurement? curious whats working for you.
check it out if you want: aisa.to
This is a strong insight because AI skill is becoming less about “who uses AI” and more about who can judge, verify, and improve AI output reliably.
The persona layer feels especially important. Raw scores are easy to ignore, but profiles like “Sceptic” or “Copy-Paster” make the gap understandable and actionable for teams, hiring, training, and internal enablement.
The one thing I’d pressure-test early is the name. aisa.to is short, but it feels a bit unclear and the .to extension adds friction if this becomes a serious AI skills assessment layer for companies. The product is already moving toward workforce intelligence, not just a small assessment tool.
Beryxa.com would fit that direction better because it sounds more like an enterprise AI skills and decision-intelligence platform. For something measuring real capability, confidence gaps, verification behavior, and team readiness, the brand needs to feel credible before the buyer even sees the scoring model.