safetyopinionbullishIf AI models are good enough to find and exploit zero-days, we should use them to find bugs first and report themVipul Ved Prakash26 Jul 2026https://x.com/vipulved/status/2081069473331548389