If you enjoy this sort of stuff make sure to support the Kiwix Project which like 90% of these commercial offshoots are based off of.

I’m reminded of that AI-written book that misidentified poisonous mushrooms.

I wish philosophy was taught a bit more seriously.

An exploration on the philosophical concepts of simulacra and eidolons would probably change the way a lot of people view LLMs and other generative AI.

The problem is that the road between creating a piece of software that does something well, and then creating simplification layers on top of it is typically much longer than just “edit a config file” and “here’s a readme”.

You need extra documentation, config gating and workflow, warnings, UI/UX work etc.

I know there are Linux elitists but kind of expecting that much extra work for what is still at it’s core mostly volunteer software seems like it’s own form of elitism.

Some additional nice things about guix:

Everything is guile. The system definition, the service definitions for shepherd, everything.

Shepherd is hands down the best init program I’ve ever used. It’s just incredibly simplistic but because it just runs the guile definition you give it, you can do some incredibly complex things that systemd etc. can do as well.

The OS documentation is built into the distro, with “info guix” you get reams of configuration information for the distro without ever needing to look it up online.

About a year and a half.

To be honest it’s not “easy” to use. The guiding principle behind mainline packages is that everything has to be built from source, so most somewhat unpopular things are missing from the mainline channels.

To use it like any other distro you’re going to need to learn how to write packages fairly quickly. Luckily the main draw of guix is the entire OS being based on guile so once you get a little under your belt you can just read the specs from other channels to see how a package is written.

Took me maybe a week to start writing guix packages.

There’s also The toybox

Guix because I love the idea behind Nix but Nixlang is the most painful language I’ve ever had to type out.

Don’t they charge per token?

So they’re also making money every time somebody says please or thank you…

Yea it’s a better position to try and take them on IP control.

I assume deals like this go out the window if they also mean Netflix Ireland can’t take Irish people to court for copyright infringement.

Hledger!

Command line, plain text files so anything can read them, and GPL!

According to the Trump that’s exactly what he wants, countries to sell off the bonds they have in exchange for super long term bonds that defer interest for like 100 years.

From my experience it’s quite the opposite, cause when something breaks in guix/nix/bazzite you basically need to know how the entire subsystem works to troubleshoot it.

You can’t just copy paste some nonsense from superuser to fix it.

You really gotta hope those 5-10% LLM interpretive miss rares don’t hit your particular case.

Also if it fucks up, the doctor is still liable for malpractice right? Or do they get to kick that ball down into the abyss of trying to get LLM companies to take responsibility for their products.

There’s been a lot more evangelism about emacs lately.

Also the fact emacs has doom and spacemacs to ease new users in while vim drops you into : and tells you to swim means that it’s easier for emacs to get new users.

In terms of getting to an exact location, the most efficient is no vehicle, walking.

Cars are less efficient, followed by busses, then probably trains, then boats, then airplanes (unless you parachute).

Cars are the least efficient in terms of moving large numbers of people from places they can then walk from.

Who needs RAID when you have mergerfs

If you’re using flatpak and Firefox you have to un-sandbox the font files from both flatpak and firefox’s content sandbox

I think most ML experts (that weren’t being paid out the wazoo for saying otherwise) have been saying we’re on the tail end of the LLM technology sigma curve. (Basically treating an LLM as a stochastic index, the actual measure of training algorithm quality is query accuracy per training datum)

Even with deepseek’s methodology, you see smaller and smaller returns on training input.

He probably saw that softbank and masayoshi son were heavily investing in it and figured it was dead.

I highly recommend Obtainium to anyone who wants to keep their apps updated without needing a central report (save for the APKs that only publish on f-droid etc)