Exec: Fire Engineers Not Using AI
A hot take is sparking debate after tech exec Alex Kehr claimed any engineer not achieving ~100% of their output via AI tools like Copilot or Cursor should be let go. He framed it as a sign of inefficiency and a lack of curiosity, igniting a conversation about new performance standards for developers.
While the initial take is aggressive, it reflects a broader industry push. Coinbase CEO Brian Armstrong has stated that engineers at the cryptocurrency exchange have been let go for not embracing AI. He aims for 50% of the company's code to be written by AI, particularly for front-end development, to speed up processes. The debate over AI's impact on productivity is complex and data varies. One GitHub study found developers completed tasks 55% faster with Copilot, while a BairesDev study reported 23% of engineers saw a productivity boost of 50% or more. However, another study found that when working in complex, existing codebases, AI tools actually made experienced developers 19% slower, even though the developers believed they were working faster. Adoption of these tools is widespread. By July 2025, GitHub Copilot had reached 20 million users and was used by 90% of Fortune 100 companies. The tool generates an average of 46% of a user's code, with developers retaining about 88% of the suggestions in their final submissions. This high retention rate suggests the generated code is often considered production-ready. Metrics for measuring AI's true impact are shifting away from simple lines of code. Engineering leaders are now focusing on metrics like cycle time, code churn, and defect density to assess quality. There's a recognized risk that over-reliance on AI can lead to knowledge silos and a weaker understanding of the underlying code, making debugging more difficult. The choice between tools like GitHub Copilot and Cursor often comes down to workflow. Copilot excels at rapid, in-line code completion within a familiar editor. Cursor, built as an "AI-native" editor, offers a deeper understanding of the entire codebase, making it powerful for complex refactoring and tasks requiring broad context. However, blindly trusting AI-generated code presents significant risks. These tools can produce code with security vulnerabilities, use deprecated libraries, or introduce subtle bugs. A Stanford study noted that AI coding tools have generated insecure code in lab settings, highlighting the continued need for expert human review. Some analysts suggest that citing AI as a reason for layoffs is a form of "AI-washing," where companies use the technology as a convenient justification for traditional cost-cutting or to manage optics after over-hiring. Many executives admit their AI-related cuts are in anticipation of future efficiency gains that have not yet been realized.
