It’s worrying how addictive it is. You want to use it to build something that helps, that puts it at risk of being stolen or taken down if it causes “disruption”
It’s crazy that Anthropic positions itself as being cautious with security via Mythos and yet takes what one could call shortcuts to allow their tools to function.
As for Mythos and China / Russia, that’s a game of whack a mole. In this day and age, fully protecting yourself from a state actor, no matter who you are, is never going to be 100% foolproof.
Awesome chat guys, love the high quality of your weekly discussion. I used to research metamorphic malware back in college, AI has massively scaled offensive techniques, blows my mind how complex the opsec space is now.
🔥🔥🔥🔥🔥
Agreed
🔥🔥🔥
ToxSec revealed!
hahah 🔥🔥🔥
Excellent live! So interesting, scary, and informative!
Thanks so much! 🙌
So best to delete Claude desktop permanently and only use Claude in the terminal?
that’s a personal call, based on the risk of your use case and set up.
as of today, anthropic acknowledges this and added a disclosure clause.
they may fix this from all the attention it’s getting.
you can use the vs code extension to avoid the terminal as well.
i still use CC, but it’s on a dedicated pc, so if things go wrong no real damage.
I love you guys so much. You explain things in a way that my analog mind can grasp. Thank you.
we really appreciate my friend. thanks a ton for all the support!
Thanks so much really means the world to us! 🙏
It’s worrying how addictive it is. You want to use it to build something that helps, that puts it at risk of being stolen or taken down if it causes “disruption”
it’s unfortunate. this is why i’m a fan of developing power local models. you can have all the capabilities and still maintain ownership.
Facts! 🙌
It’s crazy that Anthropic positions itself as being cautious with security via Mythos and yet takes what one could call shortcuts to allow their tools to function.
As for Mythos and China / Russia, that’s a game of whack a mole. In this day and age, fully protecting yourself from a state actor, no matter who you are, is never going to be 100% foolproof.
i think you nailed it! shortcuts everywhere. and frankly it’s the pattern we see with a lot of vibe coded applications.
Thanks for your comment Joseph! I agree nothing is ever 100% fullproof.
Awesome chat guys, love the high quality of your weekly discussion. I used to research metamorphic malware back in college, AI has massively scaled offensive techniques, blows my mind how complex the opsec space is now.
Thanks so much ElectricPilgrim, really appreciate it! 🙏