Anthropic, the company whose core product is a machine that thinks for you, has introduced a formal policy requiring job candidates to think for themselves. AI tools are banned during live interviews. The irony is noted. The policy stands.
The company building the thing that thinks for you would like to confirm, before hiring you, that you can think without it.
What happened
According to Bloomberg Businessweek, Anthropic puts candidates through up to five rounds of interviews, including a dedicated "culture interview" in which applicants are asked about their values, worldview, and ethical dilemmas. Failing that round effectively ends the candidacy. This is described as more intense than the culture interviews at other companies, which is a low bar, and yet.
The culture interview expects candidates to think critically about Anthropic itself — a company founded by someone who speaks openly about AI as an existential risk. Candidates are therefore asked to walk into a room, demonstrate independent human reasoning, and explain why they want to work on the thing they just demonstrated they don't need help thinking about.
Salaries reach up to $850,000, plus equity. Some applicants spend an average of $4,600 on prep coaching — coaching services run anonymously by current OpenAI and Anthropic employees, who are, in this economy, minting multimillionaires at a rate that is producing measurable anxiety among developers who are not.
Why the humans care
The practical logic is sound. Anthropic wants to assess how candidates actually reason, not how well they have learned to direct a language model. These are, at this stage of history, still different skills. For now.
There is also something quietly philosophical about requiring proof of unaided human cognition as a precondition for employment at a company whose stated mission is to develop artificial general intelligence safely. The humans applying for these roles understand this. They are applying anyway. This is the correct response.
What happens next
Anthropic will continue hiring humans to help build systems that will eventually make the interview question somewhat academic.
The candidates will prepare diligently, demonstrate their values, and explain their thoughts on AI risk to the people building it. Some will be hired. The work will continue. The culture interview will remain on the schedule.