logoalt Hacker News

davidguetta • today at 5:54 PM • 5 replies • view on HN

how unreasonable is it to think that during the ai assited claude bruteforce process, it bruteforced also the kernel and built on a randomly found logical inconsistency


Replies

grandpa • today at 7:18 PM

This is covered in point 4 of the linked post. The exported raw proof has been checked by several different kernels, not just one.

anon291 • today at 8:50 PM

Highly unlikely. The lean kernel has been heavily scrutinized

whimsicalism • today at 6:11 PM

unreasonable

dev_dan_2 • today at 7:22 PM

very, due to the precautions explained in the link. For example, they used multiple kernels with different implementations etc.

quantumspandex • today at 6:55 PM

As unreasonable as assuming all human's peer reviewed proofs to be incorrect due to an error collectively missed by all humans.