Anthropic Hiring Manager Drops Coding Interviews: Claude Is Better

Why Anthropic stopped testing candidates on code

Because the company's own model is better at it than the people applying.

Felix Rieseberg, an engineering manager at Anthropic, put it bluntly in a blog post: "This month, I laid the traditional coding interview to rest." As a hiring manager, he stopped asking candidates to complete coding challenges.

His reasoning is hard to argue with inside Anthropic. "There's no more point in giving you a little React challenge than there is in quizzing people about assembly; Claude is now better at both," he wrote. If the tool your company ships can out-code the applicant, the exercise stops measuring anything useful about the person.

That is a striking admission from the company selling the model. It also flips a decade of hiring orthodoxy, where whiteboard algorithms and take-home React tasks were the default filter for engineering roles.

What replaces the coding interview at Anthropic

Portfolios, prototypes and evidence of impact — not leetcode.

Rieseberg is not giving up on hiring juniors. He is looking for a different signal. He says he scans resumes and portfolios for "little prototypes that exhibit taste, empathy, and principled thinking" — projects candidates built with AI and then refined.

The message to junior engineers is that hunger beats tenure. He points to people who joined companies and became agents of change quickly, and suggests a weekend spent prompting and finessing a working prototype says more than a timed algorithm test.

He admits the new format is not fully solved. "I'm probably not the only manager trying to figure out how to interview for people who build well with AI. I haven't quite gotten there yet," he wrote. That honesty is rare in hiring announcements, and it signals the whole industry is still improvising.

The bigger shift in tech hiring

Coding tests are becoming a proxy for a skill AI already has, so companies are hunting for judgement instead.

Anthropic's move lands in a year when AI agents have started finishing real engineering work. Our own coverage of GPT-5.6 Sol beating Factorio showed agents completing tasks that used to require a human at a keyboard. If agents can ship code, interviews that test code stop separating strong candidates from weak ones.

That does not mean coding skill is worthless. It means the scarce skill is now deciding what to build, evaluating what the model produces, and owning the outcome. Rieseberg's list — taste, empathy, principled thinking — is exactly the part of engineering that is hardest to automate and hardest to test in an hour.

Expect other AI labs to follow, and expect candidates to feel the whiplash. A generation trained on leetcode now has to build a portfolio that shows judgement, which is a much fuzzier target.

What it means for you

For candidates in the US, UK and EU, the practical change is that a polished side project built with Claude or another model may now carry more weight than a perfect take-home test. Anthropic has not published a formal new interview process, so treat Rieseberg's post as one manager's approach rather than company-wide policy.

For employers, the warning is that coding screens are getting weaker as a filter. If your interview still measures who can write a React component fastest, you are testing a skill your own tools already have.

Anthropic's hiring manager has done what most companies are still debating: he retired the coding interview. Whether the rest of the industry follows depends on whether portfolios and prototypes prove to be a better filter than algorithms. For now, the signal is clear — the ability to build with AI is becoming the interview.