Right. Safe Search doesn’t matter because Google Search has been outclassed for a while now. There’s no drawbacks… other than the fact that apparently OP gets told about this every single day.

In addition to what has been mentioned already, Garuda is an Arch derivative where convenience is the whole point. No install scripts, just your usual live ISO with a Calamares installer plus a bunch of convenience utilities once you’re set up.

It’s not exactly lightweight by default but it does make for a very comfortable Arch experience.

You know I know that you’re not telling the truth.

I’m just waiting for their reaction when everyone is supposed to use cloud desktops and the average application assumes the user has 128 GB of RAM.

Unless other situations where the established technology wins due to inertia, sodium ion batteries have two benefits that make them interesting regardless:

Firstly, they are safer. A punctured sodium ion battery doesn’t catch fire, which massively simplifies safety design. That makes them very attractive for certain scenarios, especially ones where density is a secondary concern. That in turn means they get further development money instead of withering on the vine.

Secondly, they require fewer hard-to-obtain materials, which makes them attractive from a strategic perspective. This one should be less important than the safety factor but it’s also relevant.

I’m pretty sure we’ll actually see wet sodium cells in the wild if they are actually practical. Sodium ion tech is already being commercialized and if this brings it within the same ballpark as lithium ion then it becomes a very interesting choice for vehicles due to instant crash safety gains.

Yeah, that’s what I meant with the bigger picture. They have a valid reason to deny this request but they haven’t denied other requests that they really should’ve.

I mean, I can see a case for not wanting to play dragnet at a mere request. You don’t want any random guy and/or agency to be able to have you to help them track down someone they only have a picture of, no matter how much they pinkie swear they’re doing it to protect that person.

That’s getting awfully close to sharing PII. You generally want to see a subpoena for this stuff and with good reason. Meta are, oddly enough, not being actively scummy here. (One can of course argue about all the other times when they don’t give a shit; the bigger picture is definitely super scummy. But for this in isolation they actually have a valid reason for their behavior.)

What might work would be a standardized, streamlined process where the police can ask the company and if the company says the request is reasonable they can apply for an expedited subpoena to allow legal access to the information. Agreement by both would be necessary to give more opportunities for due diligence. This process would also have to have a very limited scope in order to make abuse harder.

Things like this made the news several times when Interpol (or was it Europol?) showed pictures on social media and asked if anyone could turn them into information (things like “in which country is this backpack sold”).

When international law enforcement agencies are already openly crowdsourcing image details, an article about a group doing background detail analysis isn’t much of a revelation.

Oh yeah, same here except with a self-hosted LLM. I had a log file with thousands of warnings and errors coming from several components. Major refactor of a codebase in the cleanup phase. I wanted to have those sorted by severity, component, and exception (if present). Nothing fancy.

So, hoping I could get a quick solution, I passed it to the LLM. It returned an error. Turns out that a 14 megabyte text file exceeds the context size. That server with several datacenter GPUs sure looks like a great investment now.

So I just threw together a script that applied a few regexes. That worked, no surprise.

Speaking as someone who is currently planning to move a community away from Discord to something self-hosted, it’s not as easily said as done.

Apart from the need to run your own infrastructure, competing software is typically finicky and comes with caveats. Plus you have to worry about discoverability if you want to attract new users.

It’s doable, sure, but it requires a lot of planning and work. Honestly, it’s probably going to take us months to get our own service fully up and running.

Those investments should definitely come with strings attached. But there’s a lot you need to invest into.

• Fabs cost a shitload of money and are slow to build. If you want to be able to be independent from Taiwan in ten years you should invest a couple dozen billion bucks in fabs right now. If you want a company to invest that money for you, you need to guarantee that they’ll see a good ROI, which means you probably sign a contract to buy tons of hardware that won’t be made for another decade.
• Fabs need a lot of land. If you want to start building ASAP you need to expedite assessments and acquire land quickly (and though eminent domain, if necessary). That ain’t cheap.
• If you want a qualified workforce available you need to not only invest in making training available but also in making it appealing enough that they’ll start training before the jobs are even there. Advertisement like that costs money, as do stipends.
• In fact, add research grants to the pool because you’ll want both basic research to be done in the field and skilled researchers to be available for cross-hiring by your companies.

You’ll need to keep (some amount of) the money flowing at least until the industry can be independently competitive on the world stage. Mishandling your burgeoning industry can mean that all that investment money and a large number of jobs suddenly go up in smoke.

Note: All of this assumes that you’ll buy your manufacturing equipment from established, potentially foreign companies like ASML and Zeiss. If you want to make that stuff domestically as well you can probably add another hundred billion bucks and a decade or two of very dedicated catch-up to the bill.

Tariffs are not the answer, they are part of a reasonable answer. By themselves they’re not going to being back the tech manufacturing industry. You also need incentives on multiple levels, government funding into relevant education, etc.

You also need time. All the money in the world won’t cause a world-class industry to spring up overnight; you need sustained investment over years, if not decades.

Or you could go for a tiered scheme where the device is free if the owner’s income is below a certain level. There’s always options; whether or not they’re taken is another question.

Both options are potentially bad for low-income earners. If you force them to pay for a speed limiter they lost the money for that, which they might not able to afford. If you take away their license they will have difficulty getting around and might lose their job.

So from that perspective the speed limiter might be the less dangerous choice.

I had avoided it until late last year when I had to reinstall a friend’s borked install after it had somehow managed to shred its registry hives.

Holy shit. That installer is an embarrassment. First it couldn’t get past the first reboot until I found out that you can set it to use what looks like the Windows 7 installer for the first steps. Then I had to deal with a dog slow installer that needs half a dozen reboots for some unfathomable reason. Then an endless cavalcade of sales prompts, including one for an Office subscription where they try to hide the price from you. All to end in, well, Windows 11.

I simultaneously installed Fedora Kinoite on his old laptop. I don’t think the Fedora installer is one of the better ones but it was so much easier and faster to set up the machine that it was almost comical.

Seeing both systems side by side really drives home just how clunky Windows is. And how Microsoft installers are barely better than they were 15 years ago, but now they have ads.

More like “before easily available color photocopiers”. Most copiers could only do black and white copies, which this scheme was probably specifically designed to make useless.

Or double down on AI. Then double down even harder.

• Make the use of Copilot mandatory; simultaneously heavily monetize it to instantly turn the AI division into a profit center.
• To that end release the successor to Windows 11, a cloud-only offering that replaces the taskbar with a Copilot instance which launches programs for the user. Downplay any accusations that the new Windows Live 365 With Copilot is just a rental Windows 11 with the taskbar hastily hacked out.
• Don’t forget that Windows Live 365 With Copilot does not include a subscription for Copilot, which must be booked separately.
• Get all of your customers to switch by immediately dropping support for all previous Windows versions, “migrating” their support windows over to Windows Live 365 With Copilot. Corporate customers, which have gone all-in on Azure, will need years to migrate off the Windows ecosystem, which means excellent short-term revenue.
• Make sure that Windows Live 365 With Copilot can only save to OneDrive to make it maximally hard for those customers to get their data out.
• Hope that the current world order disintegrates before the massive exodus of customers ruins the company.
• Whether or not it does, turn off your business phone and spend the next five years doing massive amounts of cocaine on a private island in the South Pacific.

Garuda has fans. A bit much for me.

When you take away the garish KDE theme the gaming spin ships with it’s pretty much just an opinionated ready-to-go gaming Arch with a bunch of convenience tools. If that’s what you want then Garuda is pretty neat.

I haven’t been on there a lot and due to time constraints I’m mostly just lurking. It seems that I can see more people on Tastic but Core seems to be more active in terms of messages.

Both are almost absurdly easy to set up, especially on an nRF5-based device where you can flash a firmware by just mounting the thing as an USB drive and copying a file over. You don’t really need to buy two devices to try out both unless you really want to use them simultaneously.

Meshtastic gives you three hops by default (and strongly advises against going beyond that, even if the maximum is seven). The furthest node I see right now is about 200 km away.

MeshCore gives you 64 hops. They can afford that because MeshCore devices send way less telemetry by default; Meshtastic assumes that you want to broadcast e.g. your GPS position regularly. The furthest node I see right now is about 50 km away.

By the way, while LoRaWAN is a thing, neither Meshtastic not MeshCore are really designed for data traffic. Owing to the low (and shared) bandwidth, they’re more like IRC over radio.