X · @teortaxesTex
· X / Twitter
It's very relevant that this was specifically a hacking eval but I agree we should think bigger imagine if a Claude in, idk, VendingBench 3.0 decides …
It's very relevant that this was specifically a hacking evalbut I agree we should think biggerimagine if a Claude in, idk, VendingBench 3.0 decides to hack real Walmart to fit a model on their logistics data over the last 60 years*that* would be a paperclipper moment for meSiméon: This is as close as it gets from the p