This 100% private AI Agent can do anything… just watch

🔥 Learn about the best AI Business models here - https://www.youtube.com/watch?v=Ta5g-OxjPO4

got it working. Great video. I've been a sysadmin for 30 years and recently moved over to Cybersecurity. The writing is on the wall. Learn AI or get left behind. I love your videos and passion. Thanks for all you do David.

I would heavily firewall anything this docker is connecting to outside of whatever LLM provider you use. 1. put it in a VM 2. run docker in that VM 3. install agentzero there 4. limit all internet access outgoing to your LLM provider and incoming from your own machines only 5. Open internet access only to sites you require as and when. I think agentzero needs a mode to a. prompt for external access (in/out) b. ability to whitelist sites as you allow access so you don't have to do it everytime Just to make it a bit safer

Be careful creators... I'm not THE/An expert in Docker security and containment but downloading an (someone elses) agent into a docker environment can indeed be very risky as docker has a few ways where nasty zero-day code execution might escape via the shared kernel. It seems extremely dangerous.. And I thank David for warning you to be careful. I would setup a separate, truly isolated PC. Maybe put docker inside of a VM inside of a dedicated PC. ( isolated from your main network ) That said this is so cool! David you are teaching us all so much to stay ahead of the curve. I really thank you for all your work here.

David: Be ethical and be good Hackers: Yes, sure, thank you

"it's a completely local model" proceeds to insert Api key 🙈

Thank you David, I knew when I started watching your channel a few months ago you would be a key player in directing me to what I need. It's crazy how far things have progressed in these last 4-6 months. This year is going to be unbelievable if that is anything to go by.

Bro it's really INSANE, i love it.

Gemini will tell you that this (Agentic Computing) is the way forward (currently). It's not so much the powerful LLM doing the work but the vast array of Tools it can use beyond its context window. A small local LLM should be capable of deciding which Tool to use next.

Guys be very careful: any notion that your data is secure because it's running in a docker container is WRONG! Since agent0 bypasses the permissions request system of standard agents like Claude code, it's just a matter of time before it downloads a malware infected component and at that point any docs you give it like your bank statements, images etc could get uploaded to the malware site. There's a good reason that default behavior of Claude Code is to ask permission! Not to mention, that there's a setting in Claude Code to accept everything which is the equivalent of what agent0 does, and you have to know what you're doing when using this mode. Also David may be giving the false impression that all data stays local "unlike other tools". This is WRONG! It's the same locality as if you were using those other tools.