Agents are scary!

So I’ve been playing with some Agents using LM-Studio and Hermes.
At first it took me a lot of time to understand how it works and how to actually use it, but then Hermes started doing stuff.
Learned new skills, pulled tons of python packages to do more stuff. As IT Admin, it’s both cool and scary.
Cool cuz it’s machine thinking and improving (I know it’s just code, it’s not actually smart).
Yet scary since it’s extremally f*cking unsafe!
And I haven’t gotten to the worst part yet.
At one point agent needed to elevate permissions… And used grep on my bash history to look for passwords.
I’m kinda done.