Anthropic drops traditional coding tests as AI reshapes how engineers are hired

Anthropic engineering manager Felix Rieseberg said the company removed traditional coding tests from its interview process this month, arguing that classic timed programming exercises are becoming less useful for judging real engineering ability. In his essay, "Hunger Beats Tenure: Advice for Junior Engineers," he laid out three main points: coding tests are losing signal, junior engineers can now use AI to turn ideas into working products faster, and people without seniority can gain traction by building prototypes instead of waiting for permission. He cited an internal example in which Anthony Morris and Alice Zhao, both relatively new to the company, built a prototype for a Claude Code graphical interface over a weekend, with the first public version launching about eight days later. The shift has triggered a broader debate. Critics of puzzle-style interviews say a 45-minute coding challenge often measures prior exposure to a question rather than the ability to ship maintainable production software. But replacing that system is not straightforward. Rieseberg acknowledged he has not fully figured out how to interview candidates who are especially effective at working with AI. Community responses also raised concerns that prototype-driven hiring could shift more unpaid work onto applicants and still fail to show how they reason through trade-offs, debugging, and validation.