A former data analyst at a company that made an enterprise version of Facebook reveals the mundane reality of digital surveillance. The company tracked every click, page view, and interaction through standard instrumentation libraries, collecting millions of events hourly in a database called "event properties." This data was protected not by sophisticated security but by its tedious complexity.

During a hack day, the analyst discovered that with 24 hours and basic SQL skills, anyone could access revealing data like profile views - who's looking at whose profile. This mirrors how government agencies operate: they don't need advanced AI to find patterns, just analysts willing to write 595-line queries. The real barrier to surveillance has been inconvenience, not security.

The article connects this to Anthropic's failed negotiations with the Pentagon over using AI for data analysis. The concern wasn't sophisticated hacking but AI making it trivial to cross-reference existing data trails - search histories, GPS movements, credit card transactions. When AI eliminates the tedium of data analysis, the question becomes: who won't be tempted to look?