Same for Germany. I can imagine they restrict it to markets with weaker privacy laws for now.

Fine them one million a day until the generators are done. That should motivate them a bit more.

Toasters could do that in the fifties. Crispiness is directly related to surface temperature and you can detect that with nothing fancier than a bimetal strip and some clever engineering. If you want to use modern technology you could use an IR sensor and a two cent microcontroller.

Using AI for the task suggests that they stuck a camera in there and use image processing to badly simulate an IR sensor. Less reliable than two metal strips welded together but also way more expensive.

Mind you, even relatively simple fridges can technically have firmware because these days the cheapest way to implement something like a defrost timer is with a microcontroller. Of course that firmware is so simple that it doesn’t need updates and definitely doesn’t need to be connected to the internet.

Other than that I don’t really see what a smart fridge can do that you can’t cover by sticking a cheap tablet on the fridge door. Sure, maybe a smart fridge can order butter from Amazon when it thinks that yours runs out but you can do the same with your phone and “buy it again” in seconds. Assuming that you even want to buy your groceries from an e-retailer.

It’s definitely a pointless price inflation device that makes the fridge less reliable to boot.

And their business practices as well. Bringing store exclusivity to the PC games market and paying publishers to delay launches on other stores is not something I want to support.

Problem is, if your slang is too new it looks like gibberish. Of course you could try something else that stands out. Like badly translated idioms.

But it’s probably jacket like pants. Even if you tear your ass open to make your writing look less 08/15 that doesn’t mean everything’s in butter. It’s probably only a matter of time until LLMs will copper off that style as well and your clever plan turns into a zero number. And that’s where the hare lies in the pepper: Your clarity goes over the Jordan and it ain’t itching the LLMs’ bean. It’ll get trained on your stuff anyway and the cat bites its own tail.

Maybe bad translations aren’t the solution either.

The same happens to people who write a lot and pay attention to their style, such as this gentleman. The way LLMs generate text is strongly influenced by what used to be considered a high-effort eloquent style. Varied sentence lengths, triplets, typographically correct punctuation, yada yada yada. Until recently, that stuff used to be the hallmark of professional writers (and Kenyans, apparently).

What were LLMs trained on? Professional writing, first and foremost. Books, scientific papers, op-eds, blog posts. So by default they take a polite, professional tone with all the flourishes and best practices that previously used to be de rigueur for authors and the like. So now what happens when someone used to “proper” writing writes something?

Everyone thinks it’s AI slop because that’s how LLMs talk. High effort now looks like low effort. And because tech bros immediately went on to develop “humanizing” tech, any other style will soon also look like low effort. And the same is true for art: Being able to produce highly detailed pictures with spectacular colors and shadows went from something impressive to evidence that you’re just a random jackass with a ChatGPT subscription.

You can’t really win anymore. Anything short of a live performance or a feature-length movie could feasibly have been generated by a fully automated content mill. That really sucks if you want to engage in creative expression of any kind and doubly so if you once thought you could make a living off of that.

I have to disagree on side details: OS X did have version numbers starting with 10. I used it between Panther and Yosemite and I can remember Tiger being called version 10.4; it was literally printed on the installation disc.

Yes, that means “Mac OS X 10.0” had two tens right after one another. Apple aren’t Microsoft silly when it comes to names but they can be silly.

Also, it wasn’t introduced as OS X. It was “Mac OS X” until 10.7, then “OS X” until 10.11, and “macOS” since then.

I’m pretty sure that if a judge would grant that injunction there’d be one less a week later.

This would be a great time for an injunction to force the data center to shut down all of those turbines until the matter has been resolved legally.

Of course you need to be able to pay for a lawyer who convinces a judge to pass that injunction.

A business is under no obligation to ensure that employees in underperforming departments get the star treatment. If agriculture-oriented rural towns generated as much money as grifting the stock market with AI does they’d be treated accordingly. As it is they’re a cost center and will be neglected or cut entirely.

That’s not how a country is usually run but for a business it’s very much on par.

I recently helped evaluate it for company use. My test considered of vibecoding an App that takes a Pipewire screencast and does some basic image processing on each frame. In C#, which doesn’t have great options for talking to Pipewire. I did several runs with various iterations of Claude.

On the upside, it had few problems navigating DBus to negotiate a screencast handle. So that’s one annoying API out of the way.

On the downside, all attempts at taking to libpipewire through P/Invoke failed, usually because Claude hallucinated parts of the API or set constants to incorrect values. I only got a working program when I allowed the use of a prerelease Pipewire NuGet package.

The generated code was of acceptable quality but I wouldn’t allow it into a codebase without a refactor. The code has zero consistency and one time the whole solution didn’t have any namespaces. The fact that the LLM writes and rewrites the code during a single prompt means that you can get mild spaghetti as an initial state.

Honestly, I can see it for something like rapid prototyping or implementing basic scaffolding for an annoying API. But damn is Claude bad at detail work.

I am perfectly aware that these settings aren’t very private and I usually run something a bit tighter. My point was that even a mostly vanilla setup couldn’t get past hCaptcha.

That is… a rather byzantine list of requirements to get a captcha service to work as opposed to just running a Firefox derivative with tracking protection on standard and a default-configuration uBO (which is the specific configuration that led to the 100 repetitions, not some kind of recommendation).

It’s almost every time for me. Maybe they don’t like my ad blocker or my browser’s privacy settings but it’s rare for hCaptcha to let me through after three or four repetitions.

Usually I give up after ten because of it won’t let me in by then it won’t let me in after a hundred. I tried.

Please not hCaptcha. It’s basically guaranteed to generate infinite loops.

It would be nice but someone would have to rip them out of existing products and put them into new ones. Theoretically possible as a (probably fairly janky) one-off product but unlikely. Still, similar products have been made.

I don’t think those sellers are going to have much lower prices when the bubble pops. DDR RAM prices are high because not much of it is being produced. What is being produced is HBM, which isn’t compatible with DDR and doesn’t even come on DIMMs. Even if DDR production picks up quickly after the bubble pops, it’s still going to take a while until consumer products appear again.

So while times are hard right now for home PC users, I’d expect there to be period in the near future of oversupply and cheap components.

From where? Data center equipment isn’t really suitable for the home. A used SXM module on an adapter might give you reasonably affordable compute but is completely useless for graphics. Data center memory is often HBM, which you can’t just transplant into your home PC. Getting an EDSFF SSD into your PC might work with an adapter but it’s also going to take up a lot of space.

Plus, there have been cases in the past of companies buying up and destroying the inventory of closing data centers specifically so it doesn’t end up on the market to compete with current products. That might very well repeat.

When the bubble bursts I expect to see a semi-decent supply of high-end hardware for specific use cases. If you have a space for a rack with noisy fans and a 3000 W power draw you’ll be able to build a kick-ass AI inference rig for like 2000 bucks. Or a really fast file server. But I don’t think there’s going to be much in it for people who just want 60 FPS in current games and an SSD those games fit onto. That’ll take another couple years.

I think it’ll be 2030 at the earliest until we see actually interesting consumer hardware from the usual companies. Maybe China will swoop in and deliver something worthwhile in the meantime but I’m not holding my breath.

I mean, if they got fined something like 100 days’ revenues¹ they’d change their tune right quick. European law allows for that kind of fine.

¹ One day’s revenues being calculated as the total revenues in their last fiscal year divided by 365.