The Mythos model has ushered in a new era of security attack and defense. The perspective of this article is very interesting: security is no longer a technical issue, but an economic issue (how much money are you willing to spend on defense). That is to say, this goes back to the proof of work mechanism of Bitcoin's PoW. Before the AI era, hackers relied on "intelligence, experience, and speed" to find software vulnerabilities and launch attacks. Nowadays, AI (such as Anthropic's recently released Mythos model) can identify vulnerabilities that have existed for decades and have not been discovered by humans or traditional tools (including a 27 year old bug in OpenBSD). The ability of AI to find vulnerabilities mainly relies not on "intelligence", but on pure investment and computing power - throwing millions or tens of millions of "tokens" and repeatedly testing computing resources to uncover all the vulnerabilities hidden in the code. Nowadays, when it comes to software security, it's not just about finding some expert to audit the code, but mainly about spending more money/computing power than the attacker to let AI try all possible attack methods and fix all vulnerabilities. Once the attacker feels economically unprofitable, they give up. Isn't this similar to Bitcoin's Proof of Work (PoW)? The security of Bitcoin relies on computing power/electricity, not who is smarter. In summary, Under AI, software security has evolved into a 'money burning game'. It has become necessary to use the concept of "cryptoeconomics" to approach software security, which is the mechanism of economic game theory. Otherwise, it is difficult to ensure safety. Perhaps this is just the beginning of AI sweeping everything. https://www. (dbreunig.com)/2026/04/14/cybersecurity-is-proof-of-work-now.html