The AI Code Dilemma: Anthropic’s Rise and Risks
On December 27, 2025, Boris Cherny from Anthropic tweeted that 100% of his contributions to Claude Code were AI-generated. The industry celebrated, but just three months later, a code leak revealed shocking details about the software’s quality.
Key Highlights:
- Volume & Complexity: The codebase included 64,464 lines of TypeScript and functions as long as 3,167 lines.
- Bugs in Production: One known bug wasted 250,000 API calls daily—a simple fix that went unimplemented.
- Automation Over Quality: The team prioritized speed over quality, opting to regenerate AI-generated code instead of implementing reviews.
As CEO Dario Amodei noted, initial claims of AI-written code were vague—is it volume, efficiency, or accuracy? The answer matters as Anthropic seems to embrace a trend where shortcuts replace structure and quality.
Is speed the new standard in AI development? As we pivot toward reliance on AI, we must ask: Are we sacrificing quality for velocity?
👉 Join the conversation on AI ethics and software quality. Share your thoughts!
