Anthropic manager removes coding round as Claude outperforms applicants
The decision is limited to one team, while company guidance still asks candidates not to use Claude in interviews or take-home tests unless told otherwise.
Felix Rieseberg, an engineering manager at Anthropic, has eliminated his team's coding interview round, arguing that Claude's ability to outperform most applicants on the exercises has undermined its value as a hiring tool. The change does not apply across the company.
In posts on X and his blog, Rieseberg said seeing applicants frame problems and develop solutions once provided a modest indication of their potential performance. He has yet to select an alternative assessment and is reviewing resumes and portfolios for small prototypes displaying taste, empathy and principled thinking. Google principal engineer Jaana Dogan responded that watching candidates use coding agents would give her a better sense of their skills.
Anthropic's performance team has also changed its hiring test. Its lead, Tristan Hume, wrote in January that Claude Opus 4 beat most candidates under the test's time constraint. After he raised the difficulty, Opus 4.5 equaled the strongest applicants. A third version instead features unconventional, game-like puzzles that Opus 4.5 was unable to solve.
Despite those changes, Anthropic's candidate instructions continue to ask applicants not to use Claude in live interviews or on take-home tests unless they are told otherwise.
