Blame the HDMI consortium. Bastards.

That said, I’m not sure why it’d be a deal-breaker. In 2026 this will be a low-end PC. It’s using a 2 year old laptop GPU that Valve has dumped more power into.

Because TV OEMs are the ones in the HDMI consortium.

I don’t really think it’s the same.

Micron just became like Samsung. Samsung also doesn’t have a consumer DIY market brand. Companies like Kingston or G.Skill can still buy Samsung/SK-Hynix/Micron’s RAM, there’s been no actual reduction in supply.

If Intel did the same as Micron did, it’d be more like third parties could sell the consumer stuff under their own names (say, the Corsair 5 XYZ), and Intel only sold Xeons directly.

The anger for the RAM shortage should squarely be on OpenAI - they’re the ones who bought 40% of the world’s RAM supply (and not even from Micron, mind you, just Samsung and SK-Hynix) and kicked off panic buying. Maybe throw Nvidia in there for handing them the money to do it.

I don’t like the killing of Crucial, fuck Micron for that, but OpenAI is who triggered the RAM shortage, and Micron is actually the least to blame of the big 3 RAM manufacturers for the issues we’re having.

OpenAI abruptly bought 40% of global supply, and announced it.

Other companies found out about it when OpenAI announced and thought holy shit, if we hadn’t heard of this massive deal, what else haven’t we heard of?!, and so they started panic buying.

On top of that, because of US tariffs and trade restrictions, the Chinese “B-tier” memory companies, who usually buy old machines from the big 3 (SK-Hynix, Samsung, Micron) and sell this lower spec RAM at lower margins, didn’t buy up these machines as much as they usually do. They weren’t sure they’d be able to make a profit given their lower margins, should tariffs suddenly change again or other restrictions get put in place.

On the one hand, I actually think this is a very good thing. Social media is especially damaging to children.

However:

The government says platforms must take “reasonable steps” to keep kids off their sites and use age assurance technologies, such as uploading official ID or facial/voice recognition, but they haven’t specified what technology platforms should use.

I hope the law stipulates that Meta is not allowed to keep this data, or use it for any purpose other than the verification itself. Not for training, not for building a profile on someone, nothing. Unfortunately the article doesn’t elaborate on that.

If they’re allowed to keep that data, then that needs to be addressed immediately. It’d be all kinds of fucked up.

There’s probably a distinction between a Meta account and a Facebook account.

I have experience in KDE being a bit buggy too. It’s kinda crazy how powerful it is, but I guess more “moving parts” means more breakage.

After a while, I moved away from KDE.

In fairness, it’s been more stable for me than Windows.

I haven’t used KDE Plasma since Plasma 6 came out, though. I’ve heard people say it’s a lot less janky, so maybe my experience is no longer the case. Nowadays the only interaction I have with KDE is the 0.1% of the time my steam deck spends in desktop mode while I’m updating stardew valley mods.

OpenAI bought 40% of the world’s DRAM.

They bought them as whole wafers (not finished chips!) from SK Hynix and Samsung.

Then they put them in a warehouse

All of that is confirmed, btw. The part below is my speculation:

To me, that reads as if they’re using VC money to drive up RAM prices, hoping that their competitors (who are catching up) can’t buy more RAM.

It’s so anticompetitive it’s unbelievable. And of course, normal buyers are the most fucked over.

Indeed.

I have a model running locally on my NAS that does image recognition for photos in my Immich app (think Google Photos, but private). It does a decent job and runs well on AMD integrated graphics on a Ryzen 5 3400G. I just search for [daughter’s name], and there she is.

I use Firefox’s translation feature (that also runs locally and can run on low end hardware).

My sister is blind and uses an AI assisted screen reader that works way better than what she was using before.

The issue isn’t AI/machine learning in itself, it’s this tech bro arms race. It’s them manipulating models to push agendas. It’s them shoehorning an LLM into every fucking Google query. It’s them telling companies they can fire all their staff and rely on LLMs.

A decent chunk of that is due to DDR4 production shutting down. If you look to the past you can see that DDR3 prices rose a while after the introduction of DDR4 too. In fact it got more expensive than DDR4, before vanishing completely.

Another thing driving up prices is tariffs and trade restrictions - usually when the main players like Micron, SK Hynix, or Samsung want to stop selling certain chips (say, DRAM at a certain binned frequency), they sell to Chinese manufacturers who are willing to sell slightly lower quality NAND for a lower profit margin.

But that’s not happening - the Chinese companies aren’t buying up the machines like they used to, because a tariff could easily wipe out their margins. It’s not worth the risk.

Add AI to that (not that many are using DDR4), and it makes a bad situation worse.

The AI aspect may get better soon, but the top two won’t. I don’t think you’ll be able to get new DDR4 for a good price at any point going ahead. Your best bet is to buy used if you see a reasonable deal.

God forbid someone who’s made their life tech is very excited that Torvalds has come to visit them and turned out to be a really nice guy.

Torvalds will probably be the highlight guest of his entire career, and he knows it. Of course that’s enormously exciting.

You know the silly stuff at the start was actually Torvald’s idea, right?

Linus (Sebastian) spoke about how he didn’t even get the reference but Linus (Torvalds) prompted him to go and watch Highlander (“there can only be one!!”)

You should’ve kept watching it. There’s some good stuff there.

Phones have been doing a lot of post-processing for a long time.

Tbh, most phone cameras would look crap without it. It’s something of a miracle what they can achieve with a tiny sensor and a tiny fixed lens.

For most it absolutely is viable.

Linux is great for the average person, great for experts.

It’s the “pro-sumer” people that struggle most often. They’re the ones who know windows pretty well, know what apps they want to install, and have became used to the quirks of windows. They struggle to adapt.

Most people use their laptops for web browsing, YouTube, Spotify, and basic document editing. They’d be fine with Linux. They just don’t use it because laptops are sold with Windows.

Love this.

The more I’m hearing about the Pebble Time 2, the more I’m liking it and looking forward to my delivery.

But fuck the 30 day warranty. Stuff sold in the UK is usually 6 years of cover (albeit only 5 for Scotland). 30 days is actually pathetic.

Zorin is FOSS.

The fact that there’s a pro version where you can pay for support does not break the GPL licence.

FOSS does not mean everything must be free of charge. It just means the user can access the source code and modify/share it if they wish.

What a bizarre qualifier.

You may as well complain about a kitchen by saying “but can it roast my turkey without using that oven?”

Of course it needs some form of translation layer or emulator in order to run programs from other OSes.

So it doesn’t then. Cool.

It’s based on Ubuntu LTS, that’s true. But Ubuntu backports device drivers to older (LTS) kernel versions, so the performance/hardware support is often similar/the same as using a newer kernel.

I believe they call this backporting of device drivers the “hardware enablement stack”, but I may be misremembering.

PopOS uses this, but Mint I believe is a strange one. You can get a variant of Mint that enables the hardware enablement stack, but I don’t think it’s a feature of standard Mint.

Contrary to what many people thing, Gnome is extremely modular and customisable. It’s just not really exposed in the base Desktop Environment itself.

You can do literally anything with the extension system. It’s very powerful.

That does however mean that you can easily break things, which is why by default Gnome marks extensions as unsupported when a new Gnome versions come out, until the maintainer adds a text string inside the extension that flags their extension as being validated for the new version.

You can disable the version checks, of course, and just risk it. But usually I find you don’t need to. By the time a new release comes out, the Gnome beta has been available for over a month, and the extensions have already been updated in advance.