Skills assessment that produces evidence rather than a score.
Open research on what counts as proof that a person can do the job. Engineering is where we have gone deepest, across almost two decades. AI skills in the roles that now work with AI is newer work, and we publish it as the evidence develops.
Everything here is open. No forms, no gates.
Prevention starts with assessment design, then adds layered controls. Codility turns those signals into a deterministic integrity risk level backed by reviewable evidence, so a human makes the final call.
Switch the coding agent on or off for each task, or have it slip in a bug in an interview. Claude Sonnet 5.5 and GPT-6.1 Sol reached Codility within 4 days of launch. Plus skill levels in every report.
No pieces match these filters yet.
Proof that a person can do the job, produced by doing real work, with AI where the job uses it, and scored against a documented methodology. It is what a decision can be defended with when a candidate, a regulator or a client questions it.
