Why not both? Even if you have chips that are more efficient for *some* workloads, that's unlikely to entirely cancel out the headwinds facing people trying to build datacentres on the ground - training still seems to need as many power-hungry GPUs as you can throw at it. But putting some workloads in space could help. Someone providing AI compute as a service, which SpaceX already does, would need a combination of orbital and ground-based compute, it's not one or the other.

Ah, ad hominem attacks, the sure sign of someone who has a really strong argument.

All the Elon haters keep using silly assumptions in order to poo-poo the idea of orbital data centres, however if you were honestly trying to figure out if it could work without anti-Elon bias getting in the way, then it might look a little different. You're not just going to take a standard ground-based data centre and launch it into space. You're going to look at what are the most practical workloads for space AI, what are the most efficient chips for those workloads in terms of tokens per Watt. You're going to want to run your chips as hot as possible to maximize radiator efficiency, so you're going to look at which chips can handle running at 80 or 90 Celsius. And if you're launching a million satellites, then you're probably going to build custom silicon to absolutely max out your optimisations. And so on and so forth - it's called engineering. I didn't personally buy any SpaceX stock, but it wouldn't surprise me if they make their data centre satellites work a lot better than most of the haters are predicting, and I'll find it quite funny if they do.

I assume it's the Ziroth video that's linked, I watched it yesterday. This is a little different than Theranos in that there's multiple companies involved, but yeah, fake it till you make it gone wrong once again. It sounds like at least at some point, Donut Labs genuinely believed that CT Coatings actually had a revolutionary battery tech, and would eventually be able to supply it to them, per leaked emails between the companies, and maybe the initial fakery by Donut was just trying to bridge the gap until CT Coatings delivered what they promised. However, it's also clear that as time went on, the aggressive fundraising by Donut from small investors for a product that they continued to have no proof even existed, and the continued false claims about what they actually had, became hugely problematic. Exactly who knew what when within Donut Labs and the other involved companies, and what legal thresholds may have been crossed, remains to be seen.

I remember building a bunch of Cyrix boxes for in-house use at my second job out of uni. It was an interesting time for x86-compatible processors, it seemed like the x86-compatible ecosystem was going to expand with a whole bunch of manufacturers getting in on it, some with very different approaches (remember Transmeta?). Didn't really work out that way though.

*paper

A preprint of the article is available on the arXiv (https://arxiv.org/pdf/2605.02243, for those that don't have access behind Nature's paywall.

Do Zipline have a patent on their system or something? Otherwise I don't see why everybody doesn't just copy that. Keeping the drone quite high up and then winching the package to the ground avoids several of the issues seen in these videos - the drone stays too high up to blow stuff such as other packages around, also reducing the noise experienced at ground level, and winching the package down until the line goes slack ensures a soft landing. It seems like a totally superior way to do it.

Don't forget outlook.com, the webmail service, if you need more Outlooks.

Although I think the Microsoft Store Outlook is new Outlook, but I could be wrong.

I've noticed a trend on Reddit these days, now when someone doesn't like your comment, they say it's an AI post as a way of dismissing it. It would be kind of handy at those times to be able to say "Ha, no - verified human!" I guess that may not outweigh the downsides though.

The average major movie is a bit longer than that; in many cases I wish they were only 90 minutes, but blockbusters these days are usually well over the 2 hour mark, and in many cases heading to 2 and a half hours. E.g. Avengers: Endgame - 181 minutes, Oppenheimer - 180 minutes, The Dark Knight - 152 minutes. So if they were to tend towards the epic end of the scale, as you'd kind of expect for a Game of Thrones movie, it might be more like three episodes long, almost a third of a season.

I presume that's Duncan M. Hamilton? I don't read read much fantasy. Surprised he doesn't seem to merit a wikipedia page.

Well, if you're into ReactOS, I guess it's proof that you're not in a rush for anything.

Don't complain, it gave you another chance to get some upvotes.

... Porsche's profits collapsed 99% in 2025. Feel free to not mention that when talking about how their year went though, I'm sure it's not relevant at all.