Gizmodo is just regurgitating this Financial Times article into a poor quality opinion piece. Journalism is preferred to someone ranting from an armchair IMO.
It's very simple. I host my stuff on a network with an open peering policy. If you as an ISP somehow have peering issues with that, then that's a you problem. I will not pay a ransom to some shady middleman that you decide to use because your network admins are too lazy. I will (rightfully) blame you and tell your customers to switch ISPs if they have issues.
Play stupid games, win stupid prices. Just wait until Vodafone Germany customers get slow speeds and an automated warning banner on every other website they visit. "Too big to fail" until it isn't.
This is a very reasonable approach until somehow your competitors service loads twice as fast on the Vodafone Germany network.
As a business, at that point, you're basically extorted to pay the ransom or deal with a loss of revenue. Since the ransom is most likely lower it won't take long for your other competitors to start paying it as well leaving you with an objectively worse product, irrespective of your warning banner (which lefty Linda or Gradma Garry isn't going to understand).
Grok will do mostly whatever you want. It's a tool and it's your personal responsibility to use it correctly. That's the way it should be. You don't need Sam Altman to babysit you, I hope.
As long as Elon keeps messing with Grok because the facts it uses aren't compatible with his world view [1], I'll stay away from it as far as possible.
And even leaving the political questions aside - the thing is being molded into a goon generator [2], just WTF.
> It's a tool and it's your personal responsibility to use it correctly.
Sure in theory, but this never works out in practice when applied to the populace. Replace "LLM" with other controversial things — social media, cigarettes, guns.
I'm also certain anyone outside the tech sphere who takes personal responsibility when using LLMs is not using Grok as their daily driver.
Grok Code Fast 1 is quite usable for small simple coding tasks. And quite a bit cheaper than say: Claude Sonnet 4.5
Personal experience when using it OpenRouter.
That said, I prefer bigger models even if they cost more. But I do this through GitHub Copilot at a steady 10 dollars per month. I can do way more request with it than I can with 10 dollars worth of credits at OpenRouter. I don't think Copilot is making any money, on the contrary.
Actual statistics: In 2023 there were 35.3 million commercial flights worldwide.[1] In that year, there were 66 accidents in commercial aviation worldwide, of which one fatal (9N-ANC).[2] This means that the chance of being in an accident was approx 1:535,000 (0.000187 %). The chance of getting into a fatal accident was 1:35,300,000 (0.000003 %). Per passenger the chance of fatality was approx 1:61,111,111 (0.00000164 %), with 72 fatalities among 4,400,000,000 total passengers.
In contrast, the United States saw 125,700 preventable deaths in the home in 2023.[3] The country had a population of 336,806,231 people back then.[4] This means a probability of approx 1:2,679 (0.037 %).
Wayland is the way to go. The same applies to Flatpak, Pipewire, systemD, etc. I'd say that this is obvious, and doesn't even need to be argued, to 99%+ of those who actually use the Linux desktop. The only opposition to this is a small group of decelerationists with a major, irrational aversion to change.
Having only two major options, X and Wayland, doesn't mean either one is correct. It is generally true that X needs to be replaced, but Wayland is not necessarily the replacement we need. It would be good to have more competition.
On a separate note, I think it's probably true that Wayland has significant drawbacks that preclude it from being an obvious replacement.
Everybody is pushing it and trying to convince the people who have problems with it that it's completely fine and their problems aren't important (like blind people being completely unable to use the computer).
At some point the pipewire of display will come along and we'll all forget wayland was ever a proposed solution.
Pulseaudio is both a protocol and also an implementation of that protocol. Pipewire also implements the pulseaudio protocol, hence its compatibility with all existing software.
Wayland is a protocol only. Every compositor - and there are many - implements it "from scratch". The "pipewire of display" would simply be yet another Wayland compositor. No one is going to solve the problems of Wayland in one fell swoop by releasing another Wayland compositor. What is actually happening is that problems are being gradually solved by the introduction of protocol extensions, which usually get adopted by other compositors after achieving success in one.
Everything you listed is bloated, slow, incompatible, unfinished or unstable. My system worked fine 20 years ago on far less capable hardware. Now even with high end workstations systems lag, crash or have strange behavior
We are running systemd with all bells and whistles on Raspberry Pi based 1 GB RAM systems. systemd-networkd, iwd, timers etc. The base usage barely touches 350 MiBs. Our actual application is containerized with systemd thanks to broad set of options increasing system security quite a bit. It works great.
You know what's bloated? Replacing all those functions with custom bash scripts or worse system services.
> We are running systemd with all bells and whistles on Raspberry Pi based 1 GB RAM systems. systemd-networkd, iwd, timers etc. The base usage barely touches 350 MiBs.
Er. I have Linux boxes that have 128MB of total RAM doing useful work in my house (not using systemd). This is not the win you think it is.
>systemd-networkd, iwd, timers etc. The base usage barely touches 350 MiBs.
That’s absurdly high for a headless system that’s doing nothing. There are countless millions of embedded devices doing useful work today with 1/10 the RAM. They run modern Linux just fine without the ridiculous bloatware.
My thoughts on systemd are complicated but I wasn't around at the time of systemd
Personally it is my opinion that Linux really split in two due to systemd partially because of the idea of the sheer size of systemd code
I am not saying that we always need such simplicity, but that I am merely giving an opinion that there are people who actually want to understand what they are running as their root and this sense of "control" really is so hard to get from things like system-d
System-d is also thus a little "bloated" compared to other inits which really show in systems like containers etc. where most developers if possible try to have alpine containers (I have seen this especially so much in golang/rust communities partially because golang is mostly static available and rust can be done the same too or compiled with musl pretty easily)
As such, personally, I can understand both systemd and other init systems, I feel like there are some guides which prefer using hummingbird etc. (https://github.com/comfies/tldrlfs) and I feel like for actually understanding "linux" from linux from scratch, other inits can be good.
Another minor nitpick I have of systemd is that its glibc based, Glibc has some of the weirdest complexities I have ever seen and a lot of package management in my opinion has been built around it and personally it feels like the decisions were made in a different era where different types of resources were constrained and updates weren't as widespread but now it has been a mess which is why we need so many linux distros in the first place with their opinions and package management
I genuinely prefer musl for this, So I prefer things like alpine/void in the process as well yet to me, freedom matters a lot. There should be a freedom of choice in such matters and systemd severely restricts it for many.
I feel like systemd is way too ambitious and which is why it requires glibc to be more feature complete in the first place, not sure if its a good or bad thing but I am merely stating what I feel like.
As I said, I have nothing against systemd myself but I am just giving the nuance I felt like, as I was trying to build my own linux distro trying to make it hyper compact and I came into this rabbit-hole, My philosophy almost was out of curiosity regarding what are the smallest systems which are still functionable (Hint: its tiny core linux which is an absolute pleasure although it isn't "secure" partially because they run everything as root If I remember correctly but )
>We are running systemd with all bells and whistles on Raspberry Pi based 1 GB RAM systems. systemd-networkd, iwd, timers etc. The base usage barely touches 350 MiBs
Okay but what are your thoughts on alpine, Alpine's motto or the first thing you see in bold letters on their website (https://alpinelinux.org/) is
Small. Simple. Secure.
Alpine Linux is a security-oriented, lightweight Linux distribution based on musl libc and busybox.
Combined with either gcompat to run glibc or personally I genuinely prefer golang/rust applications (mostly golang) like running gitea on alpine etc. and I found it to be an absolute pleasure server side to work with mostly, except sometimes software download especially python when I was running alpine on android using userLand was a somewhat-issue but maybe I had skill issue or something but I genuinely learnt a lot trying to install python on it.
Bun/Deno just works out of the box, in fact deno is even available in the apk format of alpine out of the box
I truly love alpine/appreciate its message. I feel like systems should be small partially because that means that such software could run even on much older systems just out of the box
Alpine features raspberry pi images and there is dietpi which has some decent low iso file sizes, Check them out as well if possible
Personally I love alpine but I also love the idea of using debian or some immutable distro which uses systemd and then running alpine in container, it seems to be the best of both worlds really.
When distributions started to use systemd, it was extremely buggy.
The first 24h of me using it, I found 3 different bugs in journald where it was losing data.
I'm currently using systemd, but it was far from being ready when all the fanboys with very basic use cases were insulting anyone who complained about it.
WTF? You can't address people like this on HN, and comments like this are completely unacceptable. You might not owe containers better, but you owe the community better if you want to participate here. Please take a moment to read the guidelines and make an effort to observe them if you want to keep participating here.
Wayland won't run fast on my n270 CPU based netbook. Pipewire it's good but sndio it's much faster. And SystemD it's a joke compared to the simple setup at /etc/rc.conf under OpenBSD and rcctl.
This has been happening for many years and they're not AI-related scrapers. These are attackers who port-scan as much IPs as possible, and attempt basic attacks on whatever they find. In your case they found Gitea, and started scraping your git commits likely in search of leaked credentials. They'll also look for Wordpress sites and try common weak credentials on wp-admin.
They're very easy to stop through honeypots and basic checks. You don't need Cloudflare. Anubis should work, but is probably more powerful than you actually need for these bots.
GNOME's libadwaita solves this beautifully. It's simple, nice looking, yet powerful. You could absolutely use it to make an ffmpeg front-end that's both fully featured and friendly to less technical users. But if your app can't, then another good option is to have a "simple mode" and "advanced mode".
And IMO, Handbreak is more complicated than CLI ffmpeg. It's really chaotic.
I'm not sure about NPM specifically, but in general: Pick a specific version and have your build system verify the known good checksum for that version. Give new packages at least 4 weeks before using them, and look at the git commits of the project, especially for lesser-known packages.
Gizmodo is just regurgitating this Financial Times article into a poor quality opinion piece. Journalism is preferred to someone ranting from an armchair IMO.
reply