The technology sector is experiencing a sharp divergence in hiring practices as artificial intelligence tools reshape technical evaluations. While many companies are adapting to AI-augmented workflows, Mercor, a ten-billion-dollar artificial intelligence training startup, is deliberately retaining traditional whiteboard interviews and prohibiting AI use during assessments. Head of Product Osvald Nitski outlined this strategy during a recent appearance on the Twenty VC podcast. Mercor has phased out take-home assignments and work trials, which have gained popularity as practical skill indicators. Instead, the company relies on whiteboarding sessions testing systems design, statistical reasoning, experimental setup, and technical judgment. Nitski emphasized that although Mercor uses AI for initial candidate screening, its full-time interview process remains intentionally low-technology. The approach addresses concerns that widespread AI adoption causes developers to outsource critical decision-making. The startup aims to verify that engineers possess genuine technical judgment and can navigate complex architectural challenges without algorithmic assistance. This philosophy contrasts with a broader industry shift. Over recent months, major firms including Google, Microsoft-owned LinkedIn, and Cisco have modified evaluation protocols to permit AI assistance during interviews. AI-native startups like Lovable, Cursor, and Cognition have similarly embraced AI-augmented assessments. Emily Cohen, Head of People and Operations at Cognition, noted that prohibiting AI tools during technical interviews resembles administering a math exam without a calculator. She argued that modern development inherently integrates AI assistants, making hiring processes reflect that reality by testing tool integration rather than raw code generation. The contrasting strategies highlight ongoing tension within the industry. As generative models lower barriers to code creation, companies are reevaluating technical competence. Mercor's stance suggests that foundational skills like systems architecture and analytical judgment remain difficult to automate and require unaided verification. Conversely, the migration toward AI-integrated assessments indicates many firms now view proficiency in prompting and debugging generated output as the new technical baseline. The outcome may influence broader recruitment standards. If Mercor successfully identifies engineers who maintain independent reasoning despite AI saturation, it could validate traditional screening for roles requiring deep architectural oversight. Meanwhile, the industry-wide adoption of AI-augmented evaluations continues redefining skill assessment, prioritizing adaptability over isolated coding proficiency. As artificial intelligence embeds itself across software development, recruitment will likely fracture between foundational verification and workflow integration, permanently altering how technical talent is evaluated.