It looks like the open hardware printer might actually be happening:

https://www.crowdsupply.com/open-tools/openprinter/updates/on-the-road-to-launch

I must say, working on Queerutils in my spare time is making me really happy. I'm hoping I can speedrun getting to the point where I can dogfood it in a few weeks. That's when the real work will begin

When I was a computer science student at university (just a few years ago), I was required to take a technology ethics course (called "tech and culture" but it was primarily about ethics).

The first thing we talked about in that class was how technologies are never value-neutral. "It's just a tool" has no meaning as far as ethics is concerned.

People making the "it's just a tool" argument are self-reporting that they have never bothered to think about the ethics of any technology. It is an admission that they just do not care about ethics.

CC @luxliquida @Alonely0

The Codeberg repo for Queerutils, the explicitly antifascist, queer-friendly, and no-LLM alternative to GNU Coreutils is now public! Find it here: https://codeberg.org/caden-parajuli/queerutils

It is in very early stages, and anyone is welcome to provide their thoughts on what they would like to see in this project

I've also set up Zulip for chat: https://queerutils.zulipchat.com. You're encouraged to join the Zulip if you're interested in the project and its mission, whether or not you plan on contributing

The codebase is in Zig. You can also contribute in C, although the scaffolding for this isn't quite there yet. I'm planning on getting that done this weekend

In my spare time I have started building an alternative to GNU coreutils. I know there are a few alternatives already but the goal of this project is to offer an explicitly antifascist, queer-friendly alternative, with no LLM contamination

The goal is to eventually replace more than just coreutils, but this is the beginning. I know I won't get there on my own, we'll need a community for this. I'm hoping to get a repo public on Codeberg soon

#Linux #NoAI

I'm going to go out on a limb and say it's highly unlikely that Anthropic hasn't already been watermarking their LLM output for a long time. I'd be shocked if any of the AI companies weren't. The difference is just whether they expose the watermarks to users.

Think about it: these companies train their models on the entire internet. Now most of that internet is the output of their models. To avoid model collapse, it's in their best interest to watermark their own output so they don't ingest it for training. It's also in their best interest to hide this from their competitors to sabotage them

Turns out a C rewrite wasn't necessary, at least for now. I was able to get a 30% reduction in peak and average memory usage by replacing a single transitive dependency.

The worst part is this dependency only implements a single simple type (ListT), does it poorly, and has over 50k downloads. Yet less than 20 stars on GitHub.

The culture of loading in 100s of packages for simple things without reading their code is unacceptable. It permeates many languages, but the worst offenders in my experience are JS (by far), Rust, and Haskell

I'm really starting to believe that language package managers were a mistake

Time to rewrite a slow Haskell dependency in C.

It's extremely intertwined with the existing codebase, so I can't really just rip out the whole thing and rewrite it. I have to incrementally refactor one piece at a time and keep a similar interface, which means lots of FFI use, so this'll be fun