[rescue] Q about sun2 boot flow

[Joshua Boyd]

On 9/5/26 5:47 PM, Mouse via rescue wrote:
>> If O.P. wanted an AI response, he'd have asked the AI chatbot
>> himself.
> Maybe (I don't know the OP).  But if it were me, well, if I wanted an
> LLM[%] response, I would have significant trouble getting it; I know
> next to nothing about what is available and how to use it - and my
> impression is that most of them are unusable without the HTTPS Web,
> which renders them nonstarters for me.

Acquiring an LLM model might be difficult without HTTPS, although I am 
not 100% certain of that.  HTTPS isn't required to run one, just a lot 
of CPU power.

To have it address something like this would probably require providing 
tool use of a compatible browser, possibly behind a proxy like SearXNG, 
in which case the browser might not need HTTPS, but SearXNG certainly would.

I'm not sure what models would realistically fit on the hardware you 
prefer.  Needle 2 45M looks like it might, but I don't know how usable 
it is on SS20 performance, nor if it does a good job with reasoning or 
tool using.

> [%] I refuse to call LLMs "AI".  For starters, actual intelligence
>      would be able to tell the difference between useful information and
>      the kind of gibberish that gets called hallucinations when produced
>      by LLMs.

Unable to tell the difference between useful information and gibberish 
can describe many a wetware neural net in addition to LLMs.