VibeCodeTube
In the future you will downvote nothing and b̵̰͇̹͔̩̹̲͉͉̟̜͂̓̊͝ȩ̸̛̹̠͈͍͓̬̥̱̰̯͎̖̤̫̏̐̍̏̐̀̍̓͜͜ ̴̢̧̰͉̗̠̼̹̙̩̱̫͖̫̂̇̃̅̎͆̕͜ͅh̴̙͓͎͉̩͙͎̪̥̺̔̈a̷̡̰̗͚̪͈̩͆̋͘p̵̢̨̙̪͚̞͇͎̱̰͕̈́̃̂̽̋̒̚͠p̶̛̛͕̓̌̒̒̅̿̃̃̿̌͠͝y̶̢̛͍̟̥̲͇̠̗̼͍͓͐̒̋̿̽̄̓̋͐͊̎͘
Why does this keep happening? I told Gemini to make no mistakes, FFS.
“You are a pro coder, you make no mistakes ever, Make me a Youtube please.”
✨
It’s like a comedy routine.
- “Question and no guessing or mistakes gemini”
- “Confident answer that’s totally made up”
- “Were you guessing or hallucinating in the last statement?”
- “I apologize, you’re 100% correct that I was guessing”
- “How do I get you to stop guessing”
- “Use this code…”
- Repeat endlessly
I do think it comes up with some buried, interesting sources, but that’s about it. It’s like a 5 year old.
It can’t not hallucinate. It’s just predicting (not even selecting) next tokens. It doesn’t know what it knows and what it doesn’t know. It can’t introspect. It just gives probabilities for all possible tokens in it’s vocabulary based on the context window and the inference engine selects the next one (based on it’s settings). Without having the correct answer in the context window it can just make a prediction based on it’s (fixed) neutral net parameters and these are severely limited, even for big models. What I mean is, they basically “learned” the whole Internet and compressed the whole thing into some hundred billion or a few trillion parameters. That’s an insane compression ratio. This compression is lossy. For niece information and the results are similar to the “unimportant” details in highly compressed JPGs, you can make out the general image but fine details are just a mush. The LLM itself doesn’t know this, it just gives wrong predictions.
For what it does I think the result is extremely impressive but the way it works is severely limited.
From what I understand from what it told me about itself, it can be wrangled further. That is what the paid for versions are. The info can be sandboxed and then other agents verify the correctness of the info from very specific, known to be solid, sources. This is very expensive and still not fool proof. Am I wrong in thinking this bubble is going to pop hard?
If you wan to know if a bubble pop just look at the fundamentals. Yes, I know, especially during bubbles people tell you that fundamentals don’t matter but they always win in the end. The thing is that you cannot bet on them because the market can always stay longer irrational than you can stay liquid. Eventually however it always corrects on the fundamentals again. Those can change of course over time but looking at the insane amounts of money flowing into data centers with no possible way of recovering that cost, I think the picture is clear. We also have wonderful highly circular money flows that to a large extend do not even exist but are all taken for full.
The only question is when it implodes. Within a year, within three? Who knows.
It’s definitely going to pop hard, because those “verifying agents” are just more models computing correlation with sources, not actually verifying anything.
It’s a text generator… that’s like telling a parrot to not repeat.
The amount of bugs in the Youtube app has been crazy lately. They’re clearly relying heavily on vibecoding.
Dude, so many times I have encountered bad software lately, and I always thought, some Idiot must have used LLM for this. And I don’t even know why. Everything that already was working good started breaking down for some reason. Why are devs wrecking their code base just to fix minor bugs?
The execs and shareholders demand it in order to justify the massive investment they’ve made in their shitty chatbots.
We need more paperclips and we need them now!!!
Apple is at least as at fault here. The new ui system they introduced about a year ago is a shitshow
New you say? Probably there isn’t enough training data for the model yet!
Isn’t it unbelievable how multi-billion-dollar corporations push out updates that should never have been released, despite the warnings from their undoubtedly highly competent developers?
That alone says everything you need to know about the times we live in…
It seems we’ve reached the end of the road for hyper-capitalism.
I mean why should they care? They continuously fuck people over, and they keep coming back. I would do the same at that point. #leave
Yes, that’s true, because of the network effect. But you can still rest on your laurels as long as there’s no serious competition. Another motivation for PeerTube & Co.
Regarding federated applications: I think they not only need content, but also have to become significantly more user-friendly to ever have a chance in the mainstream. It’s simply a reality that the average user doesn’t know the first thing about the applications they use—and, above all, that they never want to know. The essential and only “selling point” is and remains convenience—and even setting aside the lack of content, federated applications unfortunately can’t keep up. Not for technical reasons, but because the average internet user is such a complacent wimp.
Yeah. “me me me, now now now”. And they get exactly what they deserve. Convenience, at any cost.
But honestly it’s not even really convenient. It’s just accoutumance. Windows is the opposite of simple. Google is not simple (try to find anything in that google drive soup). iOS is terrible at doing anything other than giving Apple money. They are just used to it. That’s all. And I don’t think we should replicate any of this in the Fediverse. Because I’m a Fediverse user, I love how it is, and I DO NOT want to see it flooded by stupid flows taken from greedy useless silicon valley product managers. And if that’s what keeps Rando Joe from using the Fediverse, so be it. Maybe they’ll come later after getting fucked a bit too hard by big tech and realize that, yeah, maybe they can spend 10m learning something new, because fundamentaly, it’s better. We are building an alternative, not a replacement.
The A/B test produced less dislikes, therefore succeeded, enable 100%
Just had to squeeze that ai button in there
The app gets messed up on my Japanese LG TV every now and again (and I seem to have lost the ability to thumbs-up videos again), but it’s never been unusable. The app on my android phone looks fine at the moment.
I have been using YouTube Redux and the element remover tool of Ublock Origin…
Makes Youtube much more pleasant (plus makes it look like it used to before the stupid UI updates)
Is there also a trick to make it less slow to load, like old.youtube or youtube.classic or something? I mean, they have no performance goals or what?
rehike or invidious is a thing, has been quite a while since i tried it so no idea if they still work reliably tho
There are user agent switcher extensions for FireFox et. al. that can sidestep the garbage, anticompetitive throttling Google does against FireFox.
It’s also nice for those certain stupid websites that think they cannot work in FireFox…
If youre using an adblocker they intentionallly throttle service
Youtube redux just changes the CSS
Idk, try different browsers, VPNs, VPN configurations? I don’t usually have trouble with loading stuff… it takes a few seconds which is to be expected
I think its done by design to drive people away from using a browser where people have more control and the website has less control (of course depending on the skill and patience of the user)
Having a shitty service on browser and on certain devices just drives people to view Youtube more and more on self contained smart services like Smart TVs or dedicated viewing devices.
I use Linux, firefox and have ad blocking extensions … and sometimes Youtube just becomes unusable
App experience has been shitty too.
This is iOS yt app
Have you tried FreeTube?
It’s the only way I watch YouTube on PC these days and it works pretty well most of the time. Also nice since the YouTube website almost never works for me.
Grayjay is also good. Both on Android and pc with sync and subscriptions and history.
I’ve been meaning to try this. It’s backed by Louis Rossman, so I’m sure it’s great.
