Felix Rieseberg says AI has hollowed out the classic technical test and wants to hire for hunger and taste instead
The whiteboard coding interview has outlived many tech fads. At one corner of Anthropic, it has now been retired.
Felix Rieseberg, an engineering manager at the AI company, says he no longer uses traditional coding interviews when hiring for his team. The catch is that he hasn’t settled on what replaces them.
Why the test stopped working
Rieseberg laid out his reasoning in a blog post published on Tuesday, October 27, 2026. His core argument is that AI models such as Anthropic’s own Claude now outperform humans at coding tasks. If a machine can ace the exam, the exam stops separating strong candidates from weak ones.
He was candid that the old interview was never a crystal ball. It was, in his framing, a modest signal that happened to be worth something.
“Being able to quickly define problems and find elegant solutions was a weak but useful predictor for how successful you’d be on my teams,” Rieseberg wrote.
Note the past tense. A weak but useful predictor has become, in his view, a predictor that no longer does much predicting.
Rieseberg’s position carries extra weight because of his day job. He leads development of Claude Code Desktop and related tools at Anthropic. He is, in effect, one of the people building the thing that made his own interview process obsolete.
What he wants to hire for instead
Rather than leaning on technical test scores or years of accumulated tenure, Rieseberg says hiring should prioritize qualities that are harder to fake with a chatbot. At the top of his list is “hunger,” meaning raw ambition and drive.
He also values fresh perspectives. Beyond that, he wants people who can wield AI tools to ship high-impact projects quickly.
He is considering portfolios and self-initiated projects as possible yardsticks. The idea is that a body of work can reveal “taste, empathy, and principled thinking” in ways a timed puzzle cannot.
None of this is a finished system. Rieseberg is weighing these approaches, not rolling out a new rubric. For now, his team’s hiring process is somewhere between the old world and whatever comes next.
The view from inside Anthropic
Rieseberg’s stance fits a broader pattern at Anthropic, which has been experimenting with AI-assisted development tools throughout 2026. According to the research findings, the company has watched junior engineers turn out rapid prototypes this year.
That observation matters. If less experienced engineers armed with AI can build viable prototypes fast, seniority starts to look like a weaker proxy for value. It also explains why ambition rises to the top of Rieseberg’s wish list. When the tooling narrows the skill gap, the differentiator becomes who actually wants to build something and keeps pushing until it ships.
What this means for engineers and employers
For job seekers, the message from at least one Anthropic hiring manager is that grinding through practice problems may matter less than it once did. A portfolio of self-started projects could carry more weight, especially projects that show judgment about what to build and why.
For employers, Rieseberg’s post raises an uncomfortable question. If AI can pass the technical screen, what exactly is the technical screen screening for?
The harder problem is that his preferred traits are notoriously difficult to evaluate. Hunger, taste and empathy do not come with a scoring rubric. A coding test, whatever its flaws, produced an answer that was either right or wrong. Judging a portfolio for principled thinking requires more time and more subjective calls, which can introduce its own biases.
There is also a fairness wrinkle. Portfolios tend to favor people with the free time to build side projects, which not every candidate has. Any replacement method will have to grapple with that tradeoff, and Rieseberg has not claimed to have solved it.
The irony is hard to miss. AI labs are building tools that reshape how software gets written, and those same tools are now reshaping how the labs pick the people who build them. Anthropic’s products helped break the old interview, and one of its managers is now trying to work out what should come next.
Portfolios and self-directed projects are on the table, but he has yet to find a full replacement.
