Writing the most brilliant code, yet making the most basic, stupid mistakes?! Anthropic recently released a small tool called Claude Code, which allows users to chat with Claude and write code directly in a black window on their computer. However, they made a basic mistake during the packaging and release: they also included a "source code map" (a large 57MB file) in npm (a platform for programmers to publish code). This map file was originally intended for programmers to debug, but it contained all the original, easily understandable code (1906 files). So now anyone can download, unzip, and see the complete source code of the Claude Code tool in minutes. Specifically, it includes how the interface is designed, how to interact with the AI, how to process commands… everything! 😂 However, only the code of this small tool was leaked, not the "brain" of the Claude AI (the model itself). But many people think this is already explosive enough, because it allows them to directly study how they designed the tool. The company later realized something was wrong and quickly deleted the files. They also asked GitHub to delete some of the uploaded code repositories, but the code was already spread everywhere.