Codacus
Stop paying for AI you don't need.
Most YouTube AI content is sponsored takes on tools you'll be subscribing to forever. This channel goes the other way.
Codacus is about running real AI on hardware you already own — small models, self-hosted stacks, the homelab tools nobody's building thumbnails about. Every demo is on a 5-year-old MacBook or a $150 used GPU, because that's what most people actually have.
New video weekly. Sometimes twice. Subscribe if "free and local" sounds like a feature, not a compromise.
Codacus
Quick update : As per the last poll, I'm putting together a proper suite to test local models across different real-world use-cases and actually rank them. Building it out now — results soon.
In the meantime, the new video's up if you missed it: how to run DiffusionGemma locally on a 8–12GB VRAM card 👇
1 week ago | [YT] | 5
View 1 reply
Codacus
what AI rabbit hole do you want me to go down? Models, hardware, agents, the cost stuff comment down. Reading every reply. 👇
1 week ago | [YT] | 21
View 40 replies
Codacus
Whats the minimum acceptable tokens/ sec for a local llm for you to consider it usable
1 month ago | [YT] | 22
View 37 replies