Anthropic drops coding interviews after Claude exceeds candidate performance on React and assembly benchmarks

Rieseberg documented the shift from interview formats that tested rapid problem definition in live code. He observed that engineers new to frameworks but aggressive in AI tool use delivered higher impact than tenured staff with deeper library knowledge. This pattern emerged after direct observation of junior workflows at Anthropic rather than benchmark speculation.

March's 1991 paper on exploration and exploitation established that organizations require periodic influx of naive agents to surface overlooked friction points. Rieseberg supplied two concrete cases: his own failed dashboard proposal at Notion and Ji Pei's successful replication of the identical request one year later. Both incidents illustrate how accumulated internal knowledge blocks action even when external conditions remain unchanged.

The operational implication is that middle-management layers designed for expensive execution now face pressure to route unfiltered junior proposals directly to decision points. Companies retaining status-based gating will see measurable output divergence as AI removes the knowledge-accumulation bottleneck that previously justified tenure filters.

Labs adopting AI-native hiring without interview gates must revise promotion criteria within 12 months or risk misallocating headcount toward execution speed instead of outcome ownership.