Anthropic Software Engineer Interview Experience — Distributed Coding, AI Safety, and Hardware Reasoning

The author describes four interviews for an infrastructure-oriented Anthropic role. Distributed coding covered mode and median computation using supplied messaging interfaces, with attention to partitioning and communication costs. A separate programming exercise converted sampled call stacks into function events and explored recursion and limited recent-history data.

The culture discussion considered uncertain risks from AI breakthroughs, possible delays, and competitive pressure, alongside questions about complex situations and conflict. Another session used a spreadsheet-like tool to estimate computation, memory, and network costs under different model-sharding choices. The applicant was expected to identify the limiting resource and compare alternatives. Nearly a week later, the company requested references. The applicant was subsequently rejected after those contacts, with feedback that the interviews had not demonstrated enough senior-level capability. The report does not identify a confirmed role level.