It can run some smaller llms even with an integrated GPU or purely on CPU, something like llama:3.2:3b its also useful and you'll get a feel for things before you can run the larger models.

8-18b run very easily alongside other stuff,

18-30b run but I have to close other graphics and ram heavy stuff while using them.

Above 30b is a hit and miss and often gives my PC a seizure.

Depends, if you have a little experience with docker its smooth sailing

I have an RTX 4080 and it works as well as it can on consumer hardware without bankrupting yourself , I'm guessing anything over 3070 would work well

Ollama for the runner, multiple models like gemma3 , qwen2.5, mistral for different tasks, and openwebui for chatgpt like UI, there's a lot more you can do but this does like 95% of the job.

It can help people know that its you, but it won't bring back your followers from the compromised nsec

Wallet of Satoshi

Local AIs are very helpful , especially for asking stuff you wouldn't want to ask the big tech hive mind AIs

Ive been giving it medical reports from me and family and its been as accurate as the doctor in all cases, and even helps in explaining the cases further, which the doctors never have time for, or you don't think of asking when you have face to face time.

Creating weak men*

They forget what it was like without the freedoms, something about good times cresting men..

So much so I've built my complete home networking stack on it

Yggdrasil has worked for me really well, what issues did you face?