>LLMs really aren't that impressive at pentesting

The point is that Mythos apparently is quite capable and has developed novel exploits on its own.

That's the claim, yes. Has any proof been made available yet? (Genuinely asking here because I haven't been paying that close attention.)