In France we had an ex-president in prison not long ago. Because he was condemned for corruption and receiving money from Muammar Gaddafi. So yeah corruption will always be there to some extent.

The very concerning thing for the US is that this corruption at the top is getting widespread and there will be a point where it will be completely self sustaining. And the American justice system seems to be completely unable to push back on it. Even worse its starting to get corrupted itself with the “anti-weaponization fund” attempt.

Anybody found the footage ? I searched a bit but I didn’t find it on YouTube.

I suspect the AI use is the problem.

AI are very keen on recommending deep modification of your system. And then they quickly forget they asked you to recompile your own kernel based on a 2014 forum post. I’m fairly knowledgeable on Linux and if I would have naïvely listened to Deepseek I would have destroyed my system many times. And that’s even with prompts asking the AI to avoid outdated sources and to ask me for current configs.

I believe you would boot a liveUSB of cachyos, install that terrible idea of a browser and it would work straight away. Well if you ask AI on how to install it then it will probably recommend you to compile it yourself etc… Or to use some weird AUR repo… So maybe that’s the right context to actually use the Arch Wiki on an arch based distro ? But please, enough improvisation with AI, you breaking your system in a subtle but critical way is just one AI hallucination away.

So basically my advice is to try on cachyos without ever using AI. Just the wiki for cachyos, then the Arch wiki and finally the cachyos forum asking politely and in a detailed manner for help and being patient.

The very premise of all of this is still crazy though. There is no feature in Brave that would justify tossing a whole operating system for…

Samir, you’re breaking the car !

You mean using signal-goats ?

eayavas in this post :

If I’m not mistaken many recent datacenters are focused on inference rather than training. But it is often a bit of both.

Ah it is indeed yet another gog-games domain… Not to confuse with gog-games.org.

Also while searching for the .to website I just saw a report of a user suspecting a malware infection after downloading a game there.

Honestly, seeing how many websites share similar domains I would be very cautious browsing them…

But the banner is indeed real.

Edit : For more transparency, Here is the reddit post being suspicious about this site :

https://www.reddit.com/r/PiratedGames/comments/1qb0whb/goggamesto_is_safe_not/

But there is also reports that says its safe for the games but to be careful with the ads on the site that might lead to dangerous stuff. Probably a good idea to browse with a decent adblocker at least.

Edit : The banner was on a different website gog-games.to

Even on gog-games i can’t see that banner.

Neither on the official gog.

In fact I don’t see OP banner on any GOG related website. And not on any site using the same logo at the top.

Maybe it’s because we don’t use the same variant of Proton ?

My default is proton-cachyos and I usually use GE when I have issues.

Honestly I’m not really willing to use GE as a default. Proton-cachyos is, if I understand correctly, a proton version optimized specifically for my OS and is pretty much based on similar proton-experimental upstream.

For now HDR support is not really worth switching things up for me. Even on Windows honestly I didn’t see a difference on my pretty high end OLED ultrawide monitor.

I will try again later but for now if it doesn’t work fine with my DE or my proton version of choice I don’t think it’s really worth the complexification.

I read carefully this article but I think it is not about the issue I was mentioning.

I was talking about “model collapse” and this seems rather about multiple models training on similar datasets (shared learning ressources).

Isn’t AI training on itself a well known thing to avoid ? If I remember correctly the “” performance “” goes to shit very quickly when you train a model on it’s own output.

I doubt serious AI actors will make that mistake.

But on its own, the way they just opened their torrent client and started downloading made me furious.

In France you can still get caught downloading illegally and it can have serious consequences. But for AI businesses, copyright holders seem to look the other way. Businesses have extra rights to citizens and it’s completely unfair.

I suppose once the pillaging of everything they can train on is over, it will calm down a bit. But maybe I’m naïve on that front…

Lol I switched to cachyos also about 4 months ago and I completely agree with you.

My only game that doesn’t work is Battlefield 6 and it’s just because their anticheat doesn’t authenticate linux environment. They could but explicitly said that would permit linux cheaters to ruin the game. Well too bad, I lost one game and gain a lot of freedom and “sovereignty” over my operating system. I also have a dusty windows partition that I dread having to boot after 4 months. I’d rather not imagine the state of that W10 outdated partition :/

I also think I game better on cachyos mostly because I have more free RAM and less background stuff running I don’t need while I’m playing.

The only thing I don’t really agree with you is HDR. I kind of gave up for now on tweaking games to support it. On some games I never could make it work even with a lot of tweaking (gamescope etc). On some other games I made the effort but just couldn’t see the difference so it seemed not worth it. On Windows I had nothing at all to do to get 100% functional HDR on all games.

Still very very happy to be gaming on linux and the fact that I play even more on linux than I did on Windows shows how practical it is.

I worked for years on a large email infrastructure for a job and for me it’s absolutely not worth it either.

I would prefer to take a subscription on a reputable host.

Why?

Because even if I do everything perfectly at setup (TLS, SPF, DKIM, DMARC) that will still be precarious.

The security of SMTP is a patchwork of protocols added on top of it and a bunch of opaque reputation systems. If anything ever goes wrong with my email my domain’s reputation would fall. And that’s the thing, once your domain reputation goes too low, you can’t fix right away and say “my bad” and recover. Your mail will be silently blocked like Spam until a few days of sending perfectly clean emails. You need time to recover.

So mail self hosting is accepting that at any time if you make a slight mistake, your communications to other will be almost impossible for days. And again since a lot of it is reputation based you can’t fix the issue and recover immediately.

The business I was working for had everyday scenarios like that. A client that failed to update its DKIM and didn’t notice right away. When they do their reputation on for example Cisco’s platform is super low and we filter them as spam. And then it took days for them to recover even if they fixed the DKIM just one or two days after their mistake.

On the other hand I could take a protonmail subscription and use a domain that has so much volume and is tracked so carefully in term of reputation that I know my mails will be received and have all the necessary security done right.

These reputation systems are inherently difficult for small volume mail domains. There is no other users ln your domain so one mistake is all it takes to start having delivery issues and most importantly silent failed deliveries that you dont know about.

Is it possible? Yes. Is it necessary? Not really. If you can pay for a privacy respecting host…

Hence for me it’s not worth it because there are privacy respecting providers so it’s not like I absolutely have to self host it.

Now that I think about it, if the playlist are indeed stored on Google’s servers, I could probably also use google takeout to require my personnal data. It’s what pipepipe/newpipe uses to dump subscriptions from YT.

Honestly, I would prefer if a video projector wasn’t tossed as garbage if you can just buy a cheap adapter and put it in a box next to the podium.

We have enough electronic waste as it is!

Not that it matters much but isn’t there cheap adapters to/from VGA?

I think the figure was that almost half of the traffic on the internet was related to porn sites.

I have no idea if there is any updated stats for that.

I suppose you could surveil a major internet node and check all DNS requests to estimate how much traffic is “porn related” overall.

I’m not sure to follow what an NFT would bring to the table to solve this problem. I think an AI could create a NFT just as well and pretend to be human.

Also blockchains are usually pretty slow so minting a new NFT everytime you need to prove you are human would probably be unpractical.

The more I think on this problem the more it looks challenging. Almost anything digital could be forged by an AI in some way.

The only thing might be a problem only an human can reasonably solve and so it ends up like captchas. And captchas are notoriously weak to AI now.

I’m not sure there is any easy way to identify a human from a machine. If there is a way it will almost always need the human to solve a complex problem everytime which will be very annoying.