A GLITCHING AI. A MYSTERIOUS NAME. A DISCOVERY TOO STRANGE TO IGNORE. When mathematician Matthew Watkins signed on to test an early version of ChatGPT, he expected dry technical work. Instead, he uncovered a series of "rare tokens" that made the system go haywire. Some produced gibberish. Others turned hostile. One trigger word - petertodd - sent the system into what looked like full-blown paranoia. Was it a simple bug? A backdoor planted by hackers? Or a glimpse into how AI really thinks? Watkins's search for answers pulls him deep into the secretive world of AI safety research - where billionaires play god, rogue chatbots make fortunes and enemies, and the line between science and belief blurs. What starts as a programming mystery becomes something much bigger: both a window into the eerie inner worlds of LLMs, and a cryptic clue to what might be history's greatest heist. The system has since been patched. The window is closed. But for a brief moment, we saw inside - and what looked back at us was maybe more human than we'd like to admit.