Isn’t this still true for the large language models and math is detected and handled by an external tool? Anything you can give as a source for the latest state of the art?
Sending power outage context to the vehicles does not seem like enough of a response. I hope at least they have internal plans for more. For large, complex systems, you want multiple layers of protections. The response feels way too reactive when they could use this incident to guide improvements across the board.