Anthropic manager drops coding interviews as AI tools improve

The end of the coding test

Felix Rieseberg, an engineering manager at Anthropic, has decided to stop using coding challenges during his team's hiring process. He shared this decision in a blog post titled "Hunger beats tenure: Advice for junior engineers." Rieseberg oversees engineering for several major products, including Claude.ai and the Cowork tool.

The manager explained that traditional coding tests have lost their value because AI models have surpassed human capabilities in those specific areas. He noted that there is no longer a reason to give candidates a small React challenge or quiz them on assembly language. He stated, "Claude is now better at both," as reported by The Next Web.

While these small challenges were once considered a useful predictor of success, Rieseberg admits they are no longer reliable. He observed that his most impactful builders often succeed regardless of how they perform on these specific types of tests. This shift reflects a broader reality where AI can handle the rote mechanics of programming.

Prioritizing impact over tenure

Rieseberg argues that the way engineers gain success is changing because impact now depends less on accumulated knowledge of specific libraries or frameworks. This shift creates a new landscape for the job market where newer engineers can potentially overtake more experienced colleagues. He believes that the ability to deliver results is becoming more important than the number of years spent in the industry.

The manager specifically noted that he is seeing relatively new engineers fly past people who have been around much longer and are technically proficient. He describes this trend as "hunger" beating out tenure. This means that the traditional hierarchy of experience is being disrupted by the speed at which AI allows people to produce work.

For those without established social capital or power in a company, Rieseberg suggests a different path to success. He advises these engineers to identify a specific ambitious goal the company wants to achieve. Instead of focusing on learning every framework, they should focus on solving a single, meaningful problem for the organization.

The prototype method

To prove their value, Rieseberg recommends that engineers build a working prototype over a single weekend. He emphasizes that candidates should avoid trying to cram too many different ideas into one project. The goal is to show a functional solution to a specific problem rather than a broad, unfinished vision.

He provided a specific example from within Anthropic to illustrate this point. Two relatively new engineers, Anthony Morris and Alice Zhao, built a prototype graphical interface for Claude Code over one weekend. This project was then able to ship its first public version in roughly eight days.

This approach prioritizes the ability to execute and ship products quickly. By focusing on a prototype, an engineer can demonstrate their ability to take an idea from concept to reality. This method allows them to showcase their specific contributions in a way that a standardized coding test cannot capture.

New hiring signals

Because traditional tests are becoming obsolete, Rieseberg admits he has not yet figured out the perfect way to interview for people who build well with AI. For the time being, he is looking for different signals on résumés and portfolios. He scans for evidence of "taste, empathy, and principled thinking" in the prototypes candidates submit.

These qualities represent the human elements of engineering that AI cannot yet fully replicate. While an AI can generate code, a human must still decide what to build, why to build it, and how it should affect the user. Rieseberg is looking for candidates who can navigate these higher-level decisions effectively.

This shift in evaluation mirrors a broader trend in the tech industry. Other major companies are also rethinking how they judge engineering talent. For instance, Google is currently expanding pilots that allow candidates to use Gemini during their engineering interviews.

The future of engineering

The move away from coding tests highlights a significant change in the software development lifecycle. As Anthropic has noted, Claude now writes 80% of its own code. This means that the role of the human engineer is shifting from writing every line of syntax to overseeing the output of AI systems.

Amazon has also joined the conversation by warning its engineers against focusing solely on AI usage metrics. This suggests that while AI is a primary tool, companies are still trying to figure out how to measure meaningful work. The focus is moving toward the quality of the final product rather than the tools used to create it.

For the reader, this means that learning to prompt, architect, and oversee AI systems may be more valuable than memorizing syntax. The ability to identify a problem and build a prototype quickly is becoming the new standard for professional success. The industry is moving toward a model where the best builders are those who can leverage AI to achieve the most impact.