Anthropic manager drops coding interviews because Claude is better at them

Anthropic engineering manager Felix Rieseberg said he dropped coding interviews from his team’s hiring process. In a blog post titled “Hunger beats tenure: Advice for junior engineers,” he discussed the change; the report does not say it applies company-wide.

Analysis

The signal is more about changing the proxy for engineering talent than about Anthropic’s near-term earnings. If AI tools make unaided coding exercises less representative of on-the-job output, hiring may shift toward evaluating problem framing, debugging, system design, and effective use of AI. That could advantage candidates with strong learning velocity and pressure competitors—including OpenAI, Google, and Microsoft—to revisit screening practices. It could also reduce demand for conventional coding-assessment products if adopted broadly, though the article provides no evidence of industry-wide uptake.

The counter-risk is selection quality: removing a familiar filter may increase interviewer burden or admit candidates whose tool use masks weaker fundamentals. One manager’s policy is not evidence of a company-wide hiring change, let alone a measurable productivity gain. Near term, treat this as a weak talent-market signal, not a revenue catalyst. Over 1–3 months, watch for broader adoption and changes in engineering hiring or assessment practices; over 6–18 months, the economic implication depends on whether AI-assisted teams deliver more output per engineer without degrading reliability. The thesis weakens if teams retain conventional screens or report higher onboarding, defect, or attrition costs.

Key Decisions for Investors

• No direct position in ANT on this item alone: the hiring-policy signal is too narrow to establish a change in growth, margins, or valuation.
• Add to a watchlist for broader evidence that AI-assisted coding is changing hiring criteria across major software employers; corroboration would matter more than another isolated policy announcement.
• Monitor coding-assessment vendors as a conditional risk rather than a short thesis: look for customer losses or reduced bookings tied specifically to employers replacing timed coding tests.
• For any longer-term productivity thesis, seek evidence in engineering output per employee, release cadence, defect rates, and retention; worsening quality or onboarding outcomes would falsify the optimistic interpretation.