Das Claude-Code-Leak in vier Diagrammen: eine halbe Million Zeilen, drei Unfälle, 40 Werkzeuge [OC]

Data source: aggregated counts from a community mirror of @anthropic-ai/claude-code@2.1.88 (npm publication March 31, 2026)

Tools: Python / AI agent

I don’t think nearly enough people are talking about what this says about the AI companies‘ prediction that AI is going to take over white collar work. If ANTHROPIC is making mistakes this sloppy, what makes them think bigger enterprises are going to hand over their code to these tools at any real scale?

I work in a heavily regulated industry, in software, and the stakes are insanely high for us to make stupid mistakes like this (a ransomware attack basically to put a competitor out of business and ground essential processes to a halt for months). There are efficiencies, yes, but we aren’t itching to lay off our developers who are much less likely to do stuff like this.