Developers Warn AI-Written Code Breeds Flaws and De-Skilling as Google Says 75% Is Machine-Generated
Updated
Updated · 404 Media · May 13
Developers Warn AI-Written Code Breeds Flaws and De-Skilling as Google Says 75% Is Machine-Generated
5 articles · Updated · 404 Media · May 13
Developers forced to use AI coding tools say the systems often produce flawed output that takes longer to review and fix, turning promised productivity gains into extra work.
Anonymous engineers and designers described a growing loss of skill and mounting security and quality risks, saying companies are creating hard-to-audit codebases and long-term tech debt.
75% of new code at Google is now AI-generated, executives say; Microsoft has put the figure at up to 30%, while Anthropic says about 90% of most teams' code is machine-written.
Those internal adoption pushes have coincided with AI-linked job cuts rather than better products or shorter workweeks, including Meta's 10% reduction, Microsoft's 7% U.S. voluntary retirement program and Snapchat's 16% layoff.
Tech giants are firing thousands while AI creates flawed code. Are they trading human expertise for a 'rat's nest' of expensive tech debt?
With AI generating 75% of code but degrading developers' skills, who will be qualified to fix its inevitable, complex mistakes in the future?
AI Now Writes 65% of New Code at Tech Giants: Productivity Gains, Quality Risks, and the Future of Software Development
Overview
The rapid adoption of AI-generated code by major tech companies is transforming software development, with organizations like Snap now relying on AI for most new code and Google redefining engineers’ roles from code writers to editors-in-chief. This shift is accelerating development cycles and driving an industry-wide move toward machine-assisted programming. However, it also introduces new challenges, such as balancing productivity gains with risks to code quality, managing internal tensions between rapid AI deployment and code standards, and ensuring that human oversight and skills remain central as AI becomes a primary contributor to foundational coding tasks.