General activity goes out the VPN always. My self hosted setup, lab stuff, and work stuff gets different connections out (logically, not physically).

The opinions of Gnome are not a match for mine.

KDE let’s me get really specific on the things I want to, and its well supported.

Just that simple for me!

Bc everything I’ve ever read says the opposite: that people spend more time trying to figure out what went wrong when you trusted “AI” (implication: AGI… except we all are aware that these do not exist yet in anything besides the wet dreams of managers) misbehaved, like deleting the entire prod database or escaping from its container and wreaking havock, though performing “well” since for some reason it was trusted to measure its own performance, hence it simply deleted all unit tests and now it gets a perfect score, no flaws, yay!

Eh, I think the examples you have there are the penultimate examples of bad decisions by bad decision makers.

LLMs are quite good at verifying requirements of a spec, reviewing budgets to make sure items aren’t missed, etc.

That said - I agree the writeup is nonsense. LLMs dont understand goals, which is why the person is important. It also makes some common mistakes about middle management, more the caricature of shitty middle management than what good middle management actually is.

No, I don’t see LLMs replacing anyone in any effective capacity any time soon. If used as an appropriate tool, they can be beneficial. What we have today is NAI (narrow, small scope and good at single purpose tasking) not GAI (general AI, sometimes noted as AGI - human level reasoning). And thats what would be needed to replace a person, but its so far beyond current capabilities its laughable when they pretend its right around the corner.

For reference, jackasses like Musk and Altman claim it’ll happen within the next year or so. Other dipshit corporate types claim its just 5-10 years away.

The reality is an LLM will never understand physical or social cause and effect. They still struggle with long context reasoning. This is why you’ll see the dipshits and jackasses try to redefine from general intelligence to something like replacing a desk job without fucking up, because its just marketing crap, shifting goalposts to play pretend that their claims of being geniuses are true.

They aren’t, LLMs aren’t what will get us to general AI, and if they didnt try and push their hype machine so much we’d probably have better and more focused LLM-based tools.

Anyway, a super long way of saying I agree its crap, but disagree on why.

Just about anything is OK for a server, I use whatever hardware I get from work thats being decommissioned.

The biggest limit I see is the ram. Depending on what you spin up and how, it’ll be the tightest budget per service.

Then tag with [CBH] as required.

No, but maybe its an experience thing - have to be burned enough times by not having something to want to have something?

I think a lot of projects start off just someone goofing around with an idea so no hard requirements to start. I do that when I’m learning something new but they aren’t ambitious or worth sharing.

I always start with defining requirements, and then I follow with a bunch of pseudo code. Its how I do things at work too. I like getting the ideas and structure out before I write any actual code or I get stupidly distracted.