MintsJohn

The honest answer is i was fed up and choose the nuclear option out of frustration and just wanting to try a new toy. So yeah your right, i'll try that.

I don't think the inconvenience is that big though few secs loading page per session. The ridiculousness though kinda is, we've had years and years of all manner of crawlers that managed to behave, i've even used site rippers in the dial-up days that avoided hammering servers and fetching dupes, now the ai-bros come along and tear the internet apart, replacing it with their chats systems and generic slop.

Identified AI bots that finally made me cave and put anubis (with very liberal limits) up were anthropic/openai/perplexity according to the headers they sent. They were also over 99% of traffic despite proper robots.txt (site only has a few hundredth visits per day), which they ignore. I don't mind bots or crawlers, but these bots got the same page/query string every time they decided it was time for a crawl, which was about daily, cache control, last modified headers they also never seemed to be interested in. They were very very far from well behaved, unlike search engine crawls, and impossible to tame.

And it's all measured in "intelligence", a completely meaningless term. For coding i'd be much more interested in how much context actually works, what the complexity of algorithms it can understand and create is, for what languages. How much it manages to follow existing structures or that is just adds ad-hoc machinery to pass the test, etc etc.

A smaller model in the same generation will never be the same as a bigger one, assuming this is a smaller model, and the same generation, as naming implies, it will not be comparable, it might be on the benchmarks, even on the benchmarks that matter, but the whole story should also give the drawbacks.

I tried codeberg, used it a year, then early this year in all their wisdom codeberg decided to show adverserial random text instead of my repo, reporteldly to mess up llm training to user agents they weren't sure were human.

Codeberg had one job, serve my repo, it didn't do that, when brought up, I was told it was a feature not a bug, they could maybe whitelist me but that wasn't my problem, it was that random people got totally blocked or from accessing the repo. I moved back to github.

Glm5.1 is fantastic for me. But that could be how I use it, I don't ask it to build entire apps or entire features, instead asking it to build piecemeal functionality. For that it compares very well to chatgpt 5.4 (I haven't extensively tried 5.5, it might be better, might be same). I have given deepseekv4 pro a try but not much more than a try, as it performed subpar on 4 tasks in a row (missing the obvious/intended path, generating subpar slightly buggy code to make things work the not obvious way) , I gave up on it.

Glm5.1 for me was a bit of a llama3.1 moment (first open model i could chat with that was usable in manging my inputs the intended way) for code, the first open model that was actually usable.

Of course the devil is in the details. What you say and the skills needed make sense. It's unfortunately also the easiest aspects to dismiss either under pressure as there is often little immediate payoff, or because it's simply the hard part.

My experience with llms in general is that sadly, they're mostly good bullshitters. (current google search is the epitome of worthlessness, the AI summary so hard tries to make things balanced, that it just dreams up and exaggerates pros en cons for most queries). In a same way platforms like perplexity are worthless, they seem utterly unable to assign the proper value to sources they gather.

Of course that doesn't stop me from using llms where they're useful; it's nice to be able to give the architecture for a solution and let the llm fill the gaps than to code the entire thing by hand. And code-completion in general is a beautiful thing (sadly not a thing where much focus is on these days, most is on getting the llm create complete solutions while i would be delighted by even better code completion)

Still all in all, the more i see llms used (or the more i see (what i assume) well willing people copy/paste llm generated responses in favor of handwritten responses) on so much of the internet, resulting in a huge decline of factualness and reproducibility (in he sense, that original sources get obscured), but an increase of nice full sntences and proper grammar, the more i'm inclined to belief that in the foreseeable future llm's aren't a net positive.

(in a way it's also a perfect storm, the last decade education unprioritised teaching skills that would matter especially for dealing with AI and started to educate for use of tools instead of educate general principles. The product of education became labourers for a specific job instead of higher abstract level reasoning in a general area of expertise)

This is what finetuning has been all about since stable diffusion 1.5 and especially SDXL. And even something StabilityAI base models excelled at in the open weights category. (Midjourney has always been the champion, but proprietary)

Sadly with SAI going effectively bankrupt things changed, their rushed 3.0 model was broken beyond repair and the later 3.5 just unfinished or something (the api version is remarkably better), gens full of errors and artifacts even though the good ones looked great. It turned out hard to finetune as well.

In the mean time flux got released, but that model can be fried (as in one concept trained in) but not finetuned (this krea flux is not based on the open weights flux). Add to that that as models got bigger training/finetuning now costs an arm and a leg, so here we are, a year after flux got released a good finetune is celebrated as the next new thing :)

Interesting notion, I notice the same with image models, less stylistic more blandness on the latest generation. Only MJ seems to have style as a feature.

I tried it with rust, it's so bad it's simply not usable, it hallucinates methods and even the syntax is wrong at some points (it especially can't get error types correct or seems). Gpt4 doesn't handle rust perfect either, but the code it produces is good enough to only need some touch-ups, it can explain and fix wrong use of (we all known) libraries and even gets async code. But it's especially great for boilerplate, saves so must typing.

I was hoping openai/gpt4 would see some healthy competition, but Gemini doesn't seem to be it. Of course, the rust language might be an edge case.

No worries, you're not alone. I can see it has decent (not great imho) production values, but it's not for me. It's slow and uninteresting, I'm level 2/3 and facing unbeatable level 5 enemies (a hag in a swamp, some tiny island after a swamp or some underground minotaurs). I know I should find entertainment in doing other things in the game, but I'm 10 hours in already, and progress is agonizingly slow. All I got is that I've bug in my eye I want to get out of it, somehow all my companions have it as well, and there's a refugee camp that's being evicted by more druids. I have no idea what the overarching story is, I just go from little set piece to new little set piece (which feel rather artificial), but it's not very interesting and my characters feel as weak as when I started the game. I have fond memories of bg1 and bg2, played them into early morning cause I wanted to know what was next. But now, I just don't care, I just feel lost and like making no progress at all.

That there seemingly is no great story shouldn't be that much an issue to me, it doesn't always matter to me, but with the battles not being engaging, character progress nonexistent, exploring not being exciting nor rewarding, there's just nothing that makes me want to come back. Now i do have a great dislike for open-world games, and it seems bg3 just has all the elements that makes me dislike it, too much make your own adventure. I don't recall the earlier baldurs gate games being this way.

There's quite a few hosted SDXL platforms (mage.space, leonardo.ai, novel.ai, tensor.art, invoke.ai to name a few) and most consumers do not have the GPUs needed to run those models, only enthusiasts do.

It's always baffled me that stability didn't offer a competitive UI platform to use their models with, clipdrop is just bad quality and very bare-bones, and dreamstudio is pricey and still lacks most features. So this move to a new licensing strategy doesn't surprise me, it actually is somewhat comforting, as i expecting them to just stop releasing further trained models (e.g sdxl1.1 and up), and only offer those on their services (of course, that can still happen) cause how else were they going to monetize the consumers (i know they (planned to) offer custom trained/finetuned models to big corps, but that doesn't monetize consumers).

However, as most releases by stability these days, it has this feeling of close-but-no-cigar, and the recent LCM lora's might be a little slower, but these actually offer 1024^2 resolution, work with any existing lora's and finetunes (so they are usable for iterative development, unlike this turbo model, cause well, it's a different model, can't iterate on it then expect sdxl (with lora's, to a lesser extend also without) to generate a similar image) and support cfg-scale (and therefor negative prompts / prompt weighting). I suppose there's some niche market where you need all the speed you can get, but unless there's a giant leap in (temporal) consistency, that will remain niche, i don't see the mentioned real-time 3d "skinning" neither the video img-to-img (frame-to-frame) gimmicks take off with current quality and lack of flexibility. It's good research, optimizations have lots of value, but it needs quality as well.

Their recent video model is quite bad as well, especially compared to pika and runway gen-2, but well, but as with the the dalle-3 comparison one can say those are closed source and stability's offering is open.

Then we have the 3d model, close sourced, worse than luma's genie unfortunately.

The music model is nothing like suno's chirp (which might be multiple models, bark and a music model) used together), and the less said about their llm offerings the better.

Bottom line, stability needs a killer model again, they started strong with stable diffusion 1.5, took a wrong turn with 2.0 (kind of recovered by 2.1, but the damage was done), and while SDXL is't bad in a vacuum, neither was it the leap ahead that put it in front of competition like midjourney at the time, and Dalle-3 a little later, and now even a relatively small model like pixart-alpha, also opensource, can offer similar quality to what sdxl offers (with a lot of caveats, as it has been trained on so few images it just doesn't have info on many concepts). And more worrying, there's no hint of something better in the stability's pipeline. But maybe image-gen is as best as stability can get it, and they think they can make an impact pivoting in another direction or multiple directiobs, but currently, it feels a master-of-none situation.

Well, this just goes to show not all elderly are the same. My relatives would love more remotes and hate more strange non physical interfaces. Controlling lights with voice commands vs a (eventually on a remote) button, button wins 100% of the time.

The biggest hurdle seems to be discoverability. A physical remote makes sense, and doesn't change. Buttons have a singular function and context doesn't matter. Apps are different beasts, navigate up/down/forward/backward (thus context (what did you do before, now doing x does y, but otherwise z)) is just met with glazy eyes. Going back and front just makes no sense, why sometimes you need to go to a menu and other times it's a shortcut button, makes no sense, especially when summoning the menu needs special navigation. To add insult to injury, every now and then apps get an overhaul, and suddenly navigation and buttons changed/looked different.

Now when i say apps, i do mean apps on phones/tvs. The windows UI works for them because the basics are the same for all programs, using word or outlook, menu items have text like "send" "save" and while it takes time, functionality is discoverable. But for tv apps the logic is "click up or down until the icon you want has a different hue, then press a button on the remote to do stuff, but only when the screen shows x,not when it shows z". It's too much functionality condensed in too little UI. I'm constantly baffled by design choices for apps that are supposed to be used by everyone, I'm sure it looks nice to designers and devs, but have they even tried showing their brand new TV interface (these are the worst offenders) to an elderly and gave them simple tasks?