Latest 30 Comments

While I realise the morality of models trained on stolen code is as yet, unsettled, to me that's one of the smallest issues. From a pragmatic point of view, we're all trained on stolen code when we study and gain experience as developers. For me, unethical use of AI would be the specifics of how it's used. For example, re-licensing copyleft software is genuinely easy now, and a huge threat to open source. I think vibe-coding is a fast way to an unmaintainable project that nosedives after a month or so, but as an aid to an experienced programmer who thoroughly understands everything that is being output it will be a boon, only enhancing the project. While I understand you could say 'How are we supposed to vet responsible use, how about we just say no across the board?', this developer is demonstrably proficient with this codebase already, so what's the big deal?

View this comment - View article - View full comments - Report this comment

If you already understand the problems on capitalism, but then use the tools of a capitalist company that will atrophy your programming skills and basically take away your means of production, then I don't know what to tell you anymore...

View this comment - View article - View full comments - Report this comment

Quoting: TheSHEEEPCan we get an "S for Slop" rating? 😂

We definitely need a rating like "not suitable for human consumption" ;-)

View this comment - View article - View full comments - Report this comment

Can we get an "S for Slop" rating? 😂

View this comment - View article - View full comments - Report this comment

Quoting: Cloversheen

Quoting: F.Ultraso the payouts is actually divided per how much your music was played.

How much of this is "in theory"?
Because that seems like an (in reality) impossible task outside planned organised events like a music festival.

"We spent half an hour trying to figure it out, which burned through all the money"

View this comment - View article - View full comments - Report this comment

Quoting: mrdeathjr​This d7vk is very impressive with d3d3 titles and others in my case:

MotoRacer 2 DX3

External Media: You need to be logged in to view this.

StarWars Shadows of Empire DX5

External Media: You need to be logged in to view this.

Radeon ARK Demo DX7

External Media: You need to be logged in to view this.

​😀

I might have mentioned it before, but I always enjoy looking at your screenshots and seeing GKrellM. Haven't used it in ages but man, that is good nostalgia. 😃

View this comment - View article - View full comments - Report this comment

Quoting: Linux_RocksWould running with a rifle be safer than running with scissors? 🤔

Generally people aren't shooting at you when running with scissors. 😁

View this comment - View article - View full comments - Report this comment

I wonder how many old Steam games this could enable to be playable and/or verified.

It would be interesting to see the stats on how many games use each version of Direct 3D. Google says around 870 games use versions v2 - v7

View this comment - View article - View full comments - Report this comment

Quoting: Pikolo

Quoting: PhiladelphusI did notice that video backgrounds were a thing recently while video-chatting with my brother. Nice to see that finally happen.

What does doing "\@everyone" actually do? It's unclear from the patch note.

It was working the same as @everyone (mentioning everyone giving them a notification), but without warning you that you're sending a very broad notification

Thanks! I'm just still unclear what an "escaped mention" is, or at least what it's supposed to do. If "\@everyone" does the same as "@everyone" now (if I understand the fix)…what's the point of the backslash? 🤔 Or does it do something different now? What does "\@<user>" do differently from "@user"?

View this comment - View article - View full comments - Report this comment

They shold release a barebones version .

View this comment - View article - View full comments - Report this comment

I basically bolted over to my steam library to go see if MDK 2 was finally fixed after reading this, that was one of THE games I actually felt like I gave up going to linux.

The good news is: Yes, they did fix the black texturing issues in MDK 2 HD! Yay!

The bad news is: Everything regarding cutscenes continuing is broken, like everything renders correctly (except the HUD) but the dialogue doesn't continue and menu music plays where it shouldn't. Don't know if this is an issue with the game or an issue with proton, but either way MDK 2 HD is still unplayable.

Also, MDK 2 Classic has not benefitted from the Mesa patches as far as I'm aware, as it still starts up with a black screen.

One day... One day it will work. We're at least closer.

View this comment - View article - View full comments - Report this comment

Still little surprised they couldn't get proper FSR4 (or FSR next-gen) happening in these boxes.

That FSR3 upscaler is going to look VERY UGLY unless they pull off some sort of miracle update to improve it allot

View this comment - View article - View full comments - Report this comment

Meh, results matter. But also if stolen code ends up appearing; you don't want that happening.

There is a risk when using AI coding tools where eventually it could turn into bloatware. So hopefully that isn't what will end up happening a couple years down the line.

View this comment - View article - View full comments - Report this comment

This game is terrible. I tried to play it.

View this comment - View article - View full comments - Report this comment

I never liked Lutris, Heroic worked way better for me anyways. But I never wanted to be negative about it. But now I can just say it's AI slop.

View this comment - View article - View full comments - Report this comment

Quoting: Purple Library Guy

Quoting: scaine

Quoting: BladePupperThe cards you get are used to play the actual game whereas the skins from say CS2 are basically only used to cash in and maybe pay for a steam game.

Agree with most of your post except this bit. MtG packs are not "necessary" to play the game. You buy the base pack and you're good to go.

That is . . . true in some sort of technical sense. But it has nothing to do with the actual experience of anyone playing Magic.

Yeah, I know. I spent hundreds of pounds building my various MtG decks back when it first launched. I just wanted to point out that technically (definitely technically) you don't need booster packs to play the game.

Now where's my Serra Angel...

https://www.cardmarket.com/en/Magic/Products/Singles/Alpha/Serra-Angel

View this comment - View article - View full comments - Report this comment

I use a local LLM model to help plan out stuff I'm going to code or if I run into an issue I can't solve and need help figuring out a way to implement it that I might not have considered, but I avoid agentic AI like Claude Code as I find more often than not it just ends up producing garbage as you run out of context very quickly even with the big cloud models.

I have no problem if someone wants to use it in there application, but you have to be open about it and be willing to deal with the fact some folks won't use your program and will look down on your code for obvious reasons.

Also, I find that very quickly you can get to the point where the AI has generated so much code that even an experience programmer can lose track of how the code base interacts with everything, especially in the case of something like Lutris with so many moving parts (Wine, DVK, etc).

View this comment - View article - View full comments - Report this comment

My concern is the AI's were probably trained on code without respect to their licenses. So tagging the code is kind of important for legal purposes in case the landscape around it changes.

I personally don't use AI to generate code, but I might sometimes use it as a fuzzy search or use it for language or visual processing functions. Reason being that what ever I gain in efficiency, I also lose in opportunities to learn and improve as a developer.

Setting that aside code architecture is extremely important in regards to long term maintenance, robustness and security. My experience is that AI is just generally terrible at that, as I had them fail to architect something extremely basic and small correctly after nearly 8 hours of prompting across a lot of the top performing ones.

They seem to only excel at smaller functions or things that are easily templateable, which makes sense when you understand that they just duplicate what they have been trained on.

There are plenty of reasons why I favor self-hosting AI's and don't fund these corporations, the ethical issues are endless. But overall I just find the developers response here reason enough to not trust the long-term health of the project going forward.

View this comment - View article - View full comments - Report this comment

...we all know that nothing is going to improve with the current US administration.

I see, so it's somehow Trump's fault that this developer started using AI and feels the need to obfuscate it after facing criticism.

View this comment - View article - View full comments - Report this comment

I've been a Patreon supporter of Lutris since 2019. That ended the minute I read his reply. I'm instead going to spend that money on some other project. One that doesn't use generative AI...

View this comment - View article - View full comments - Report this comment

Quoting: Linux_RocksWould running with a rifle be safer than running with scissors? 🤔

Neither is safe. Running with spoons is safe. Anyone wanna do a Egg-and-spoon race video game?

View this comment - View article - View full comments - Report this comment

Imo he is not wrong. As an experienced dev, especially getting assist by claude pro is really helpful. He knows what he wants, he prompts it, he doesn't have to type it. It saves an insane amount of time. Reading code and verifying it is what you requested is way less time consuming than typing it on your own.

Most people's stance against LLM generated code is because many people have tried it, created stuff with it, which was bad because they were not experienced devs. LLMs are a new tool, you need to learn how to use it like in any other field. If you master it, it helps. Experienced devs master it way easier and better than beginners.

I'm specifically not saying AI, as there is no intelligence. They don't "think". They are just good at giving the impression they are intelligent or think. They are just incredible good at guessing and have an enormous data pool to get their info from and do their guesses on.

And claude is impressively good at coding.

I also think he should not hide it. But I can understand his reasons. Social pressure is real. So what to choose? Getting depressed because people are not well enough informed and are targeting you, or deciding to keep your mental peace? This is a personal thing, not everybody has the power to withstand the pressure of so many people.

View this comment - View article - View full comments - Report this comment

I haven't used Lutris for sometime now because I just prefer Steam and Heroic but I surely have nothing against it.
Let's wait and see if the launcher is going to start falling apart because of bad AI-generated code and if the project is going to survive after what the maintainer did.

View this comment - View article - View full comments - Report this comment

Can you even truly claim it's open source when it's using AI generated code?

Of course it still is.

What you can not tell is if it violates OTHER projects copyrights / copyleft the LLM was trained on. It's more of a legal minefield rather than it not being open source. For example if it [the LLM] spilled out GPL-3 code they would have to attribute this. But since it is LLM code the developers themselves don't know.

So in any occasion IF some one can find "their" code in Lutirs they can probably sue them.

I can understand them [the developer]. Especially for an FOSS project this big. Users often times just request random features or bright up issues one singular person is barely able to fix.

If we're being honest most FOSS projects are maintained by like one or maybe two people. While tons of others made like one commit and never show up to further support the project. Neither by bug fixing, testing nor financially but expect the core devs to do so.

So yes there is no issue in using LLMs to make that two core developers more productive and keep up with the pressure. While entitled users argue about LLM code and "cancel" the developers?

As long as Lutris overall does not decrease in quality, which is often the pitfall by people NOT knowing how to code and just vibe in the hopes it works, there is no issue here.

Also we don't know how many PRs by other peoples which are merged with Lutris are actual LLM code themselves. Yet ppl hate on Mathieu?

PS.: I would not call this "AI" as none of these tools are truly intelligent.

View this comment - View article - View full comments - Report this comment

The developer of Lutris has always been a jerk, so I just use Heroic + Steam and forget that Lutris exists

View this comment - View article - View full comments - Report this comment

Quoting: memvirus

Quoting: Ehvis

Quoting: CaldathrasInteresting. I was forced to 580 from the 550 series. Wasn't given a choice. Based on the link below, Nvidia no longer supports the 550 series, by the way:

https://www.nvidia.com/en-us/drivers/unix/

To do that I had to drop my kernel to 6.14 (which is now also unsupported I think) and remove one line from the code. But any drive above that messes up my VR (among other things), so it has to be done.

hey man, I'm trying since weeks to get back to 565 from currently 580
It updated automatically and it seems like I have no choice or way back....

I'm on POP OS (No wayland) and kernel 6.17.9-76061709-generic

Is there actually a way to get back to 565.77? Can you please tell me how?

Not exactly and it'll take a bit of doing. But I installed the old packages from the Launchpad PPA. This failed because the glue module wouldn't compile. Then I just went through the compile log and tried to fix the errors. Doing that for a "too old" driver on a "too new" kernel was just too much, so I enabled the older kernel from before my distupgrade. This combo only required commenting out a single line from the driver to make it compile.

There's not exactly a manual for this and if it's still broken after a reboot then you'll have to continue from the console. So if you're not prepared to just do it and deal with whatever happens, it may be easier to just downgrade the whole system.

View this comment - View article - View full comments - Report this comment

This is not the first time he is in a controversy because of how he handles critique (valid or otherwise), and he pretty much react the same way each time.

View this comment - View article - View full comments - Report this comment

Everyone who things he should not reduce his work with using AI: Step in and do this work, or STFU!

View this comment - View article - View full comments - Report this comment

Is there anything like lutris that has installers for old games?

I like heroic but I like to use lutris for old abandonware games because it has community scripts that automatically handle patches, fixes ect

View this comment - View article - View full comments - Report this comment

🎠

View this comment - View article - View full comments - Report this comment