Not by default. Anthropic's candidate guidance, last updated 10 July 2025, asks candidates to do take homes without Claude and to use no AI in live interviews unless Anthropic says otherwise. Applicants may use Claude to refine a first draft they wrote themselves. One performance engineering take home explicitly allows AI.
- AI in interviews
- Not allowed Not allowed unless stated
- Where
- Application materials: draft yourself, then refine with Claude. Take homes and live interviews: no AI unless stated.
- What is graded
- For the take home that allows AI, the result of an optimisation task under a time limit.
- Basis
- In the company's own words
What else is known?
In January 2026 Tristan Hume described how Claude kept catching up with that take home. By May 2025, he wrote, more than half of candidates would have scored better by handing the task to Claude Code, and a later model, Claude Opus 4.5, matched the best human score within two hours. The test is now on its third version, built from unusual, tightly constrained puzzles with no debugging tools, so it simulates novel work rather than familiar work.
More than 1,000 candidates have taken it, according to the same post, and dozens of them were hired.
When did it change?
- 10 July 2025: Candidate AI guidance last updated
- 21 January 2026: Anthropic Engineering describes three versions of the AI allowed take home
Sources
- Anthropic: guidance on candidates' use of AI
- Anthropic Engineering: Tristan Hume on redesigning the performance take home
Last checked: .