What Claude Code's Source Revealed About AI Engineering Culture

10 hours ago

Anthropic's lead engineer claimed 100% of Claude Code contributions were AI-written, leading to 259 pull requests, 497 commits, and 40,000 lines added.
A leak revealed 512,000 lines of code, including a 3,167-line single function, regex for sentiment analysis, and a documented bug wasting 250,000 API calls daily.
Anthropic's leadership repeatedly escalated claims of AI-written code percentages, from 70-90% to 100%, without clarifying metrics, creating hype.
The leaked code showed poor engineering practices: massive files, excessive nesting, and automated bot-driven issue management closing valid complaints.
The response to the leak emphasized speeding up processes and using AI to check AI-generated code, rather than adding reviews or tests.
The culture of prioritizing speed over quality, masking technical debt with resources, raises concerns about sustainability and industry standards.
The author reflects on the loss of quality as a foundational value in software engineering, considering a pivot to security or traditional trades.
Commenters note that this automated 'garbage in, garbage out' culture exacerbates existing issues of technical debt and overhyped products.

Hasty Briefsbeta