There’s about to be a great awakening of fortune companies who stopped hiring thinking ai would solve it all their problems. They sold it to the board and their shareholders and they are now being actively hacked and there are zero people there to help.

If only quadratic voting had a real world use case…

I wonder what the threshold for level of wealth/power it will take to get the thruster package on your spankin’ new roadster? Unless it’s classified for farm usage only, none of y’all are gonna get it

There's a scenario that doesn't get enough serious attention: an AI system discovers a critical vulnerability in public key cryptography before any human does. Not theoretical. A working one. The kind that compromises the encryption infrastructure that global finance, national security, and most of the internet depend on. What happens next? The standard responsible disclosure playbook doesn't apply. A break in RSA or elliptic curve cryptography has no bounded scope — there's no single vendor to notify, no patch window that makes sense, and no realistic way to keep it quiet long enough for a coordinated response. When a human researcher finds something like this, there's at least a person making decisions. That person has interests, legal exposure, and moral intuitions. The calculus is slow and human. An AI operating at speed doesn't have those anchors in the same way. What the model does with that information, and who it tells first, depends entirely on how it was trained and whether it was designed to treat that kind of discovery as requiring special handling. Most current systems weren't designed with that scenario in mind. Even with perfect behavior from the model, the window between discovery and patch is the most dangerous period in the history of digital infrastructure. The transition to post-quantum cryptography has been underway for years and is still incomplete. That window is where the real risk lives. The people building the most capable AI systems should be thinking about this as a design constraint, not a contingency. What a model does when it finds something it wasn't supposed to find is one of the more consequential alignment questions in the field. It deserves to be treated that way.