Engineering Hiring at Scale

Rethinking take-home assignments in the age of AI

Take-home coding assignments worked when candidate evaluation conditions were stable. Asymmetric AI availability has changed those conditions structurally — the candidate using Cursor with Claude 4 produces dramatically different output than the candidate using GitHub Copilot, which differs from the candidate working without AI assistance. The structural question is no longer take-home vs live coding but what assessment formats produce reliable signal given what candidates now have access to.