DESIGNING AI-RESISTANT TECHNICAL EVALUATIONS

Summary from the TLDR newsletter

Evaluating technical candidates is becoming harder as AI capabilities improve. Each new model forces companies to redesign tests. This post describes one company's original take-home design, how each Claude model defeated it, and the increasingly unusual approaches the company has had to take to ensure its tests stay ahead of top models' capabilities. As work evolves alongside models, companies need to develop increasingly creative ways to find strong engineers.