Anthropic: A Framework Replaces Code Testing with Prototyping

Felix Rieseberg, the technical lead at Anthropic and pilot of Claude.ai and Claude Code Desktop, claims to have moved past coding interviews. He emphasizes the importance of concrete prototypes and argues that AI renders traditional exercises obsolete. An internal example cites two engineers who built a Claude Code interface over a weekend, with a public version released just eight days later.

At Anthropic, prototypes shipped in days and a new hiring criterion

Two engineers at Anthropic, Anthony Morris and Alice Zhao, who were relatively new at the time, created the graphical interface for Claude Code in a weekend. According to Felix Rieseberg, the first public version followed about eight days later. He now looks for small prototypes in application files that demonstrate taste, empathy, and principled thinking. He asserts that traditional coding interviews will not return. Rieseberg has led the development of Claude.ai and Claude Code Desktop.

Why he is abandoning tests: AI would lower execution costs

Felix Rieseberg states that he has abandoned technical interviews based on executing small exercises. He argues that AI makes these tests unnecessary and claims that Claude performs these tasks better than candidates do. He believes that a challenge in React no longer makes sense, just like a question about Assembly. According to him, the central issue for newcomers is that they are not listened to in organizations historically designed to filter execution because it was costly. He describes organizations where a few decision-makers validated ideas and where one had to climb a rigid hierarchy to gain the trust needed to build. He suggests that AI changes this equation, as execution would no longer be costly.

Advantage of newcomers: seeing the frictions ignored by veterans

Felix Rieseberg argues that newcomers perceive problems that long-time employees have stopped noticing. He cites James March, who wrote in 1991 that organizations depend on a constant flow of naive and uninformed newcomers. He also recalls that Patrick Collison explicitly encouraged new hires to point out what seems strange to them. According to Rieseberg, older colleagues no longer see certain points of friction. He claims that at Slack or Notion, long-tenured profiles struggled to adopt the perspective of an external user, while newcomers to user-oriented products bring natural empathy. He adds that engineers working on infrastructure or internal tools benefit from not knowing why a solution is supposed to be impossible.

Example at Notion: a reevaluated "impossibility"

At Notion, Felix Rieseberg indicates that he was tasked with designing an offline architecture that several considered unachievable. The data model was deemed too complex, and it was assumed that the CEO, Ivan Zhao, would reject the only viable path. He asserts that the technical obstacle was indeed real, but the founder's resistance was not. According to him, no one had asked Ivan Zhao the question, as this assumption had turned into common knowledge.

His method: focus on business value and deliver a prototype in a weekend

Felix Rieseberg relies on a formula from Patrick McKenzie, written in 2011, which states that engineers are hired to create business value, not to program things. He recommends understanding what the company aims to achieve and focusing on the parts that generate revenue, starting by knowing the goals of one's manager's manager. He advocates for choosing an ambitious idea aligned with these goals and building a functional prototype over a weekend, to avoid cycles of specifications, approvals, and internal advocacy. He advises betting on one good idea rather than hiding three good ones among 20 bad ones and warns against prototypes that simultaneously attempt a new language, a backend, and mobile applications. He believes the prototype should make the opportunity obvious to management and identify the right person to champion it. Rieseberg also states that ambitious developers can use AI to surpass more experienced colleagues, that test scores no longer predict success, and that he sees relatively new profiles outperforming seniors. He summarizes this dynamic with "hunger beats seniority" and redefines the term "junior" as the absence of status, power, or network, rather than years of experience.