There's no need for the middleman in this scheme. Instead, a much simpler solution would be:

1. Website A gives you a randomly generated $TOKEN
2. You go to Government and ask it to sign something like The person with $TOKEN is of legal age. You have to provide your ID or whatever here, but the government doesn't know who made the token.
3. You go back to website A, it checks the signature of the message and lets you through

This can be automated in some way; maybe with a browser extension or some referrer-less redirect sort of thing.

It's still fundamentally shitty though, because now the government pretty much knows that you want to watch adult stuff, it just doesn't know which adult stuff exactly.

A better (but almost impossible to implement) solution would be for the government to issue everyone a smartcard as an identity document (many countries already do, but without the following features). On that smartcard is a private key, with the corresponding public key signed by the government. The smartcard can then sign any $TOKEN with true statements about you, e.g. The person with $TOKEN is of legal age, or The person with $TOKEN is called $NAME, or The person with $TOKEN has a driving license, etc. You have to connect it to your computer in some way so the website can talk to it, but it should be trivially doable with almost any modern smartphone. This way, everyone has the ability to attest stuff about them without the government being directly involved.

The reason this won't work is because it would be quite expensive to do and would take a long while to implement.

I think this might be a joke similar to https://xkcd.com/2030/

Sure, just highlighting the similarities here.

Check the sources, that does account for training.

The 3Wh figure is pretty much "pulled out of the ass". Looking at other sources listed, it does seem true that similar-sized models use ~2-3Wh per query, amortized training included. So yeah I concede the point.

Hmm, after running an LLM locally I'm not too surprised about low energy use while querying (but even then, 3Wh is very low, I was getting like 5 Wh per query while running DeepSeek locally, that said datacenter GPUs will be more efficient for this). But also, this doesn't go into great depths about the impact of training.

The article is clearly mostly manipulative bullshit. The arguments about "incompatibilities" between uutils and coreutils being used as an "extend" strategy is just bonkers, the point of uutils is to be a 1-to-1 compatible toolset, and there's no reason to doubt the developer's intention there. Even if they do introduce some extra features, most software projects that actually matter will not be using them, because compatibility with coreutils will remain important for decades to come.

The kernel of truth hiding in there is that Rust's "preferred" licensing under MIT/Apache is indeed a problem, and it should have been GPL (or at least MPL) everywhere from the beginning, especially for libraries. This is probably the worst aspect of Rust indeed, but not enough to outweigh all the awesome technical parts of it.

I'm from Russia just for context. The playbooks are exactly the same down to the justifications ("protect the children" for destroying online privacy and anonymity, "terrorism" for arresting any unfavorable protesters).

But on the other hand, tail recursion is its own reward!

I think the parent comment meant "everyone else is worse at software engineering"

The US has pretty much never been a real democracy (other than in a roman sense, as in "power belongs to rich men").

I don't have an exact example to hand because I'm not using a custom substituter for now. I have a remnant of that in my config still, here: https://github.com/balsoft/nixos-config/blob/master/flake.nix#L5 . But it should be relatively straightforward: add nix.settings.trusted-substituters = [ "http://your-substituter/" ]; to your (client device) NixOS config (e.g. in configuration.nix); add nixConfig.extra-substituters = [ "http://your-substituter/" ]; to your config flake.nix; answer yes when prompted by nixos-rebuild, and you should be good.

As for remote builders, I don't really dig them myself. They require fully trusting all users who wish to build on them and are finicky to set up. Instead I just ssh into the build machine, build whatever I need there, and nix copy it back to my laptop. That said,

intent on modifying root ssh configs in spite of security practices

You can set up your nix-daemon to run as its own user nowadays, mitigating all issues related to root entirely.

botch use of substituters on the remote builder

Never had this issue so don't really know how to help

is awkward or incompatible with non-interactive sudo sessions

This one is pretty much unfixable due to how remote building works

works for nixos build but not to switch

You should probably use nixos-rebuild switch --use-remote-sudo and run it as your user rather than root.

I think the current best solution to this is to add your substituter to trusted-substituters and then add it to the nixConfig.extra-substituters attribute of your system config flake (or do some NIX_CONFIG hackery in your .envrc if you're not using flakes). That way if that substituter dies, you can easily disable it (by removing it from your system flake) without rebuilding the configuration.

Oh, and BTW, Nix has a --fallback option that will rebuild locally if a substituter is not available, but that will make things painfully slow, better to use the method I previously mentioned.