A hot potato: Today, many software developers spend weeks working on LeetCode issues just to get through a single round of interviews. An engineering manager at Anthropic has since stopped using this cartridge entirely. Claude, he says, is superior to most candidates when it comes to these puzzles.
This manager is Felix Rieseberg. Among other things, he led the development of Claude.ai and the Claude Code desktop app. Earlier this week he posted on X that the coding interview no longer revealed anything useful to him. That’s why he took it out of his team’s hiring loops this month.
In a lengthy blog post, Rieseberg says that watching candidates define a problem and come up with a good solution used to be a weak but real indicator of how they would perform on his teams. But as Claude became proficient in these exercises, a small React challenge began to teach him as little as a quiz about assembly.
Anthropic’s performance team encountered the same issue in their take-home test. Team leader Tristan Hume wrote in January that Claude Opus 4 surpassed most applicants by the deadline. So he made the test more difficult. Then Opus 4.5 came along and suited even the strongest candidates. Hume’s third version is now based on strange puzzle-style problems that Opus 4.5 failed to solve.
Rieseberg has not yet agreed on a replacement. Right now he’s scouring resumes and portfolios for small prototypes that demonstrate taste, empathy and principled thinking. He also asked other managers to share their better ideas with us. Jaana Dogan, chief engineer at Google, responded to
The majority of Rieseberg’s essay is aimed at young engineers. The title “Hunger beats tenure” sums up the message pretty well. He argues that because building things has become so cheap, newer engineers can now overtake veterans. As evidence, he points to two relatively new Anthropic engineers who created a working graphical user interface for Claude Code in a single weekend. The first public version was delivered about eight days later.
Most conversations around AI and new grads have been much bleaker, although frankly the hiring data is still mixed. Meanwhile, Nicole Bachaud, labor economist at ZipRecruiter, told The Register that tech employers are still looking for skills most often. In the company’s October survey of 1,500 employers, 62.5 percent of information sector workers requested formal skills testing as part of the initial application process. About 70 percent have also shifted these assessments to supervised or in-person sessions to keep AI out.
Anthropic’s own candidate advisory service continues to encourage applicants to keep Claude away from live interviews and take-home tests unless otherwise stated. Rieseberg also dropped the coding round just for his own team. Therefore, anyone applying elsewhere in the company will likely still face this.
