Anthropic Manager Declares Coding Interviews Dead — Signals a Broader Reshuffling of Who Gets Hired and How

An Anthropic engineering manager drops coding tests, arguing AI has made them useless for predicting performance. Google and Amazon are recalibrating too. The question is who benefits from the new gatekeeping criteria.

Felix Rieseberg, an engineering manager at Anthropic, announced in late October 2026 that he has removed coding interviews from his team's hiring process. His reasoning is blunt: Claude, Anthropic's own AI model, now outperforms candidates at the tasks these tests measure. The traditional whiteboard problem — once a crude but serviceable proxy for engineering ability — no longer predicts who will ship the most impactful work.

Rieseberg's observation carries a specific data point that matters: on his team, relatively junior engineers are outperforming senior colleagues with deeper technical chops and longer tenure. The differentiator, he argues, is hunger — the willingness to identify a business problem, prototype a solution over a weekend, and pitch it with conviction. He cites two recent Anthropic hires who built a graphical interface for Claude Code over a weekend; the first public release shipped roughly eight days later. The shift is not isolated to Anthropic. Google is piloting interview formats that let software engineering candidates use Gemini, evaluating AI fluency and prompting skill rather than raw coding ability. Amazon issued guidance on October 27 telling engineers to prioritize customer and business outcomes over AI usage metrics, framing the current moment as the era of the "Renaissance Developer" — someone who combines human judgment with machine intelligence to ship faster. What Rieseberg is really describing is a transfer of gatekeeping criteria. Coding interviews were imperfect, but they were at least somewhat legible: you either solved the problem or you didn't. The replacement — reviewing portfolios for "taste, empathy, and principled thinking" — is subjectively richer but far harder to audit for bias. Rieseberg himself admits he hasn't solved the evaluation problem. For now, he's looking at small prototypes and resumes, scanning for qualities that resist standardized measurement. The optimistic read is that this opens doors for people who think in products rather than algorithms — designers who code, domain experts who prototype, junior engineers unburdened by assumptions about what's possible. Rieseberg explicitly encourages newcomers to leverage their "outsider perspective" to spot broken user experiences that veterans have learned to tolerate. The pessimistic read is that "taste" and "empathy" are the kind of soft criteria that historically favor candidates from elite networks, with polished portfolios and weekend-project leisure time. A junior developer working two jobs doesn't have the same runway to build speculative prototypes. The old gatekeeping was narrow; the new gatekeeping may be narrower still, just harder to see. The deeper structural question is what happens to the pipeline that produced competent engineers in the first place. Debugging, small feature work, and code review were not just entry-level tasks — they were the apprenticeship through which engineers built mental models of complex systems. If AI handles that layer, the industry needs a new apprenticeship model it hasn't yet invented. Rieseberg's advice to juniors — find the business problem, build the prototype, pitch the idea — is sound for the ambitious few. It is not a labor market strategy for an entire generation of engineers.