An engineering manager at Anthropic, Felix Rieseberg, has removed the coding interview round from his team's hiring process, stating that the company's AI model, Claude, performs better on these puzzles than most candidates. Rieseberg wrote in a blog post that while observing candidates define problems used to be a useful predictor of performance, "a small React challenge started telling him about as little as a quiz on assembly" once Claude improved. This trend has also affected Anthropic's take-home tests; team lead Tristan Hume reported in January that Claude Opus 4 outscored most applicants within the time limit, leading to subsequent revisions of the test.
After Claude Opus 4.5 matched the strongest candidates, Hume's third version now requires puzzle-game-style problems that Opus 4.5 could not solve. For his team, Rieseberg is currently reviewing resumes and portfolios for prototypes that demonstrate "taste, empathy, and principled thinking." He noted that newer engineers can now outperform veterans because "building things has gotten so cheap."
Meanwhile, Google principal engineer Jaana Dogan suggested that watching candidates work with coding agents provides more insight into their coding skills. Despite Rieseberg's change, Anthropic's own candidate guidance advises applicants to keep Claude out of live interviews and take-home tests unless instructed otherwise, and other parts of the company may still require the coding assessment.
