Something really interesting happened last night. Claude Mythos spent 60 hours and discovered a new key recovery attack in a signature algorithm called HAWK, which was designed to resist quantum computers. Hold off on burying your cold wallet in the underground garage. This didn’t break BTC. Didn’t break ETH. Didn’t break full AES. HAWK is just a post-quantum signature candidate that hasn’t even gone live yet. But here’s the part that sends chills down your spine: Human experts reviewed it for two years. AI, armed with a bunch of agents, research papers, and computational tools, dug out a new path from the mathematical structure in just 60 hours. The attack cost for HAWK’s smallest parameters dropped from about 2^64 to 2^38. In plain terms: Today’s lock wasn’t picked. But the people designing tomorrow’s anti-theft locks just realized one of the steel plates in their blueprint is actually made of paper. In the past, when people talked about quantum threats, it always felt like something far off in the future. Now, the more immediate question has become: Quantum computers haven’t arrived yet, but AI has already started auditing the algorithms we’re “preparing to defend against quantum.” In the future, encryption security might not just need to guard against quantum computers. It might also need to guard against a group of tireless AI researchers who can read papers, write code, and run experiments. The most absurd scenario is this: Humans spend ten years perfecting a lock. AI looks at it for two days and tells you: “This screw isn’t tightened.” This doesn’t mean BTC is at risk tomorrow. But it’s a wake-up call for everyone: “Post-quantum” isn’t a get-out-of-jail-free card. Any algorithm, before it’s truly deployed, needs to survive AI’s brutal scrutiny first. So, what do you think will arrive first in the future: Quantum computers, or AI eliminating a batch of post-quantum algorithms?