Anthropic manager stops coding interviews as Claude outperforms humans

The end of coding tests

Felix Rieseberg, an engineering manager at Anthropic, recently shared a blog post detailing a significant shift in his hiring practices. He announced that he has officially stopped asking job candidates to perform traditional coding tasks during the interview process. This move marks a definitive end to a standard industry practice that has long served as a primary filter for technical talent.

r>

Rieseberg explains that the primary reason for this change is the current capabilities of Claude. He notes that the AI model is now better at coding than the humans it is being used to evaluate. Because the company's own product can perform these tasks with high proficiency, the old method of testing a candidate's ability to write a small piece of code is no longer useful.

The manager describes the situation as a shift in utility. He compares the current state of coding interviews to quizzing someone on assembly language, suggesting that such tests no longer provide a meaningful look at a candidate's true abilities. In his view, if the tool used by the company is superior at the task, the test itself loses its value as a metric for hiring.

A new way to evaluate talent

While Rieseberg is moving away from coding tests, he is not stopping the hiring process entirely. Instead, he is searching for new methods to identify engineers who can build effectively alongside artificial intelligence. He admits that he has not fully perfected this new system yet but is actively scanning for specific qualities in resumes and portfolios.

He is looking for evidence of prototypes that demonstrate a candidate's unique human attributes. Specifically, he wants to see work that exhibits taste, empathy, and principled thinking. These are qualities that an AI might struggle to replicate perfectly, and they serve as the new benchmarks for what a high-quality engineer brings to a team.

The goal is to find people who can act as active agents of change within a company. Rieseberg suggests that instead of proving they can solve a single isolated problem, candidates should show they can build a project from start to finish. This includes the ability to prompt an AI, finese the output, and deliver a functional outcome that solves a real problem.

Advice for junior engineers

For those just entering the field, Rieseberg offers a specific piece of advice regarding how to approach the current job market. He encourages junior engineers to move away from a purely technical focus and instead prioritize making a tangible impact. He believes that understanding the company and the people within it is just as important as the work itself.

The manager suggests that the new standard for a successful application might be a demonstration of a project built over a weekend using AI. Rather than a whiteboard challenge, a candidate might show a prototype they prompted and refined. This approach allows the candidate to showcase their ability to oversee a project's lifecycle rather than just writing individual lines of code.

He emphasizes that the ability to navigate these new tools is becoming the baseline. Junior engineers who can demonstrate a proactive attitude and a desire to contribute to a team's goals will have an advantage. The focus is shifting toward how a person uses the available technology to achieve a specific, high-quality result for the organization.

The logic of the shift

Rieseberg's decision reflects a broader trend where the tools of the trade are evolving faster than the methods used to hire for them. He uses a quote from actor and director Orson Welles to frame his perspective on trying new things. Welles noted that there is no confidence to equal "sheer ignorance" when trying something new, suggesting that it is only when we know a profession that we become timid or careful.

This perspective suggests that the current hiring landscape is in a state of transition. Because the product the company builds already codes better than anyone it can hire, the old rules of assessment are being discarded. It is a logical progression for a company like Anthropic to stop testing for a skill that their own software has mastered.

The industry is now facing a challenge of how to verify human competence in an automated world. As Rieseberg looks for new ways to interview, the tech world will likely see a move toward more holistic evaluations. These will likely prioritize a candidate's ability to think critically, manage complex systems, and use AI as a force multiplier for their own creativity.