Anthropic engineer declares the coding interview dead

Anthropic engineering manager Felix Rieseberg says he has stopped using traditional coding interviews, arguing AI coding tools make small coding challenges less useful; Google engineer Jaana Dogan says she would be interested in assessing how candidates use coding agents. The broader industry has not widely abandoned skills testing: a ZipRecruiter survey of 1,500 employers found 62.5% of Information-sector employers require formal initial skill tests, while 70% have switched to proctored or in-person assessments. ZipRecruiter also reported that 61.5% of Information employers see critical thinking as a primary skill lacking among candidates.

Analysis

The investable signal is not that coding tests are disappearing, but that hiring is shifting from measuring unaided code production to measuring judgment when paired with AI. If this becomes widespread, routine implementation skills may lose wage scarcity while architecture, verification, product sense, and effective agent orchestration gain value. That can widen the productivity gap between strong and weak engineering organizations—and put pressure on junior hiring and traditional technical-assessment providers—without necessarily reducing total software demand.

The near-term evidence is weak: individual comments from Anthropic and Google sit alongside employer data indicating continued, often proctored, testing. Treat this as a change in assessment format, not proof of an industry-wide hiring model pivot. For ANT and GOOG, the direct financial link is presently too indirect to underwrite an earnings or valuation change; any benefit depends on AI tools translating into faster delivery, lower engineering cost, or better product output.

Over 1–3 months, watch for repeatable changes in job postings, interview processes, and engineering headcount plans—not anecdotes. Over 6–18 months, the structural risk is a thinner entry-level pipeline: firms may get near-term efficiency but weaken the future supply of experienced engineers. A reversal would be visible if AI-assisted work raises defect/security rates or if employers return to more rigorous individual technical screens. No compelling single-name trade follows from this article alone.

Key Decisions for Investors

• No event-driven position in ANT or GOOG on this evidence; the hiring anecdotes do not establish a material revenue, margin, or valuation catalyst.
• Monitor engineering hiring mix and headcount disclosures across large software companies. A sustained decline in junior hiring alongside stable output would support a labor-cost/productivity thesis; rising rework, security incidents, or delivery delays would falsify it.
• Treat interview-assessment vendors and coding-preparation businesses as watch items, not shorts: verify their revenue exposure and customer retention before acting, since employers may change test formats rather than eliminate assessments.
• Reassess the theme over the next 1–3 months if multiple employers disclose agent-based evaluations or altered hiring criteria; otherwise, classify this as low-signal commentary rather than a sector catalyst.