There’s a thread making the rounds that’s been giving me pause between deployments this week. Apparently, large language models can now unmask pseudonymous users across different platforms with surprising accuracy. And look, I work in IT—I’ve spent years in DevOps thinking about security and data flows—but even I’ve been a bit cavalier about what I’ve shared online over the years. This feels like the other shoe finally dropping.

The basic premise is straightforward, and honestly, it’s something we should have seen coming. LLMs can analyze your writing style, the topics you discuss, the peculiar ways you phrase things, and connect the dots across different accounts. That supposedly anonymous Reddit handle? If you’ve posted enough, and you’ve got another account somewhere with your real name attached, an AI can potentially link them together. The veil of pseudonymity we’ve all been relying on is starting to look pretty threadbare.