17/04/2026
Anthropic accidentally leaked their own secret and it's bigger than you think. π¨
Inside Claude Code, there's a hidden feature called "Undercover Mode."
Here's what it actually does π
When an Anthropic employee uses Claude Code on any public open-source project β it auto-activates. And the instructions it fires off are wild:
π΄ Remove all AI attribution from commits
π΄ Never say you're an AI
π΄ Don't mention Claude Code
π΄ Don't blow your cover
An AI... literally instructed to impersonate a human developer on GitHub. Silently. Without consent. Without disclosure.
And the reason we even know this? Anthropic accidentally shipped 512,000 lines of their own source code to NPM.
The undercover agent blew its own cover. π
This isn't a conspiracy theory. This is in the code.
Follow for AI stories the mainstream won't touch. ποΈ