Post

I agree, but I worry: What happens when a Mythos-level AI, without guardrails, is freely available? What are the immediate consequeces? Proprietary software is still riddled with "security-by-obscurity". Mythos-level decompilation capabilities will expose all that at once. Are we ready for that ?

Sounds like their problem for poor prioritization and inadequate defensive posturing. They can git gud. Mythos being closed source doesn't help defenders, it just fattens frontier lab wallets and increases inequity between actors.

Maybe I am overestimating the probability and/or damage of the bad stuff happening. And maybe not.

Ultimately, open source acts as a deterrent not by hiding capabilities, but by ensuring that the defense is crowded, collaborative, and inherently more adaptable than the offense

And we might be sitting on a time-bomb of legacy systems that run proprietary code, riddled with major vulnerabilities, slow to adapt, but still being relied on widely.