Existing models can already do what Claude Mythos did... #ai #mythos #claude

Rumor has it, Anthropic Mithes is scared that the whole internet might explode because their model is just so good at finding software vulnerabilities. The bad news is the current publicly available models can already find similar vulnerabilities just like Claude Mithes. So, I guess the good news is the internet is still here. So, we're fine. So, this company has just successfully reproduced Anthropic's Mithes results using publicly available models like GPT-5. 4 and Opus-4. 6. Like, just look at these results. Like, all these things that was discovered by Anthropic's Mithes, these two models can also find them. It just felt it failed in some parts where it only discovered the vulnerabilities partially, but it's still pretty good. And they found even more vulnerabilities with just using GPT-5. 4 and Opus-4. 6. And just look at the amount of zero-day exploits they found in the Linux kernel. So, it is totally not a new thing. It has always been here. It's just people didn't really search hard enough because look at the cost. Like, for one file search, it cost 30 bucks. Like, who has the money to do that? Other than like top AI labs or like VC-funded companies. So, we're probably going to be fine. And if you want to learn more about the technical side of AI, then check out intuitiveai. academy.