rozodru

there’s no point. I tried the free trial of youtube premium cause i’m on Qutebrowser and the adblocking isn’t the greatest. even with Premium i’d still randomly get ads every now and then. IF i were paying for it I’d be pissed.

Now I circumvent this by just opening all the videos in MPV.

As someone who has to deal with LLMs/AI daily in my work in order to fix the messes they create, this tracks.

AI’s sole purpose is to provide you a positive solution. That’s it. Now that positive solution doesn’t even need to be accurate or even exist. It’s built to provide a positive “right” solution without taking the steps to get to that “right” solution thus the majority of the time that solution is going to be a hallucination.

you see it all the time. you can ask it something tech related and in order to get to that positive right solution it’ll hallucinate libraries that don’t exist, or programs that don’t even do what it claims they do. Because logically to the LLM this is the positive right solution WITHOUT utilizing any steps to confirm that this solution even exists.

So in the case of OPs post I can see it happening. They told the LLM they wanted analytics for 3 months and rather than take the steps to get to an accurate solution it ignored said steps and decided to provide positive solution.

Don’t use AI/LLMs for your day to day problem solving. you’re wasting your time. OpenAI, Anthropic, Google, etc have all programmed these things to provide you with “positive” solutions so you’ll keep using them. they just hope you’re not savvy enough to call out their LLM’s when they’re clearly and frequently wrong.

with the amount of times Claude hallucinates solutions I wouldn’t trust it to write my grocery shopping list let alone national defense.

Kid in high school already wants in on the right wing grift. annnnnnnd he’s done it. the other Nazi’s will hold him in high regard like that fat dough pissy boy rottonhouse. He’ll be getting calls from Erika asking him what he thinks of her new leather pants cause we know she likes em young.

good jorb.

Brody King.

yeah I remember a year ago when they said the same. They keep moving the goal posts because they HOPE it’ll happen, but it’s not going to. LLM’s across the board, regardless of the company, have progressively become worse. They’ve sunk billions into something that simply isn’t improving and they either don’t know how to improve it or are just praying to whatever gods they believe in that it will some how magically get good.

There’s nothing left to feed these things. For example you can go to Claude right now and ask it something so painfully simple that there’s a strong chance it’ll hallucinate a solution. not even a strong chance, a guaranteed chance. Like I just asked it “With the Fish Shell how can I get cat <<EOF > notes.txt to work?” it didn’t know how. it presented 3 options all of which were wrong. It claimed I was using an older version of Fish that doesn’t support Heredocs. that’s not true. Fish doesn’t support Heredocs at all, period, and never has.

pretty much. you’ll see it first hand with the vibe coding shit. sure the basics will work but none of it will scale, it’ll be full of exploits, and just garbage all around. But most people simply don’t know better and trust whatever the LLM spits out.

I mean people hail Claude as the best out there but if you know any better and you’ve spent any time with it then you’d know it’s 100% useless. not a single solution it spits out these days is correct. and Claude Code has become noticeably worse.

We’ve known about the Housing First initiative the world over for years. Housing First with needs based assistance works, it’s proven to work. we know it works. Addiction and Mental Health issues are symptoms of being homeless.

We know the solution to the “complex” issue of homelessness. We know how to fix it. There’s just no money to be made in doing so. So we won’t fix it even though we can.

that’s fine. ChatGPT is 100 times easier to quit and forget about than Facebook. ChatGPT serves zero benefit for most people in their daily lives. It’s extremely easy to stop using a tool that constantly hallucinates solutions.

I use DOOM Emacs. is it potentially a bit of overkill? sure. but it’s got all the plugins I need, built in terminal, built in git tui, built in file manager, and depending on what i’m writing web dev stuff in it automatically sets up a sort of basic bootstrap for me.

Yes there is a bit of a learning curve but honestly I just don’t bother using anything else now.

Kate is also great.

sure they’re leaving but…where? what’s a viable alternative?

I mean if I had it my way everyone would just go back to IRC, I still use IRC. it’s great, nothings changed, it’s perfect. combine that with like mumble or teamspeak and you’re good to go.

I checked out Stoat lastnight and it’s very slow. took well over an hour to get a verification email (I assume because many others were signing up) and the current server offerings are minimal. Also a good majority of the themes listed simply don’t work. it’s growing and I’d say it’s the best alternative right now but it’s not there yet and I worry it or the various instances of it could hold up if there was a massive Discord exodus.

Matrix is a no and xmpp ain’t much better.

potentially. it’s going to become more of the norm as more nations enforce stuff like this to “protect the kids” coughget your id and know what you’re doing onlinecough

Also keep in mind Discord is just about to launch their IPO in March so the timing lines up.

they do. desperately. it’s been stated for awhile now.

Canada also opened one alongside France.

I don’t know what it’s like in Houston but here in Toronto the entire food delivery industry would pretty much come to a stand still without e-bikes. do they kill people? no. are they dangerous? I’m sorry but with the way some of those dudes drive those things on sidewalks and the amount of times i’ve seen them hit people…yeah. We’ve also had a few literally blow up on commuter trains to the point where we out right banned them on trains.

There’s more to this that I don’t think people realize. no they don’t kill people but yes they do cause accidents because of the people driving them and feeling like they can ride them on sidewalks or…you know…the batteries will randomly explode.