Old Fucking Punk @lwriemen@librem.one

I wonder how much productivity is being lost by people using LLMs to write long things where the meaningful content remains very small in comparison.

I've noticed that looking up how to do $THING with a command-line $TOOL now almost always gives me an LLM-generated page with pages of boilerplate nonsense (what is $TOOL? How to install $TOOL on Ubuntu, how to install $TOOL on macOS, and so on), with the actual two sentences of content right at the end. These are obviously generated to provide more space for ads, but there's a lot of this cropping up in other contexts.

Saving a few seconds of writing time in exchange for wasting a few minutes of reading time for each of your readers is a staggering drop in overall efficiency.

Tech Bros on A.I.'s Artificial Inevitability. https://www.gocomics.com/brian-mcfadden/2026/04/27

“In this work, we conduct a large-scale simulation of how users might delegate work to LLMs across 52 professional domains. We find that current LLMs are unreliable delegates: even frontier models corrupt an average of 25% of document content over long workflows, with sparse but severe errors that silently compound over time.”

Good to see the issue addressed explicitly, even though the results aren’t surprising—why would anyone expect LLMs to be reliable!?

https://arxiv.org/abs/2604.15597

Hi! We are Koumbit, a self-managed, not-for-profit collective based in Montreal/Tio'tia:ke. We aim to be an ethical and human alternative to big tech. We've been offering web hosting, web dev, and sysadmin services since 2004!

We work with open source software, and prioritize technological autonomy and data privacy. Our servers are in Montreal and belong to us.

If you are looking for local, socially conscious, ecological and ethical web hosting, please think of us!

@signalapp It would probably help if Signal itself didn't use what looks like a real conversation or story to communicate to the user. It legitimizes phishing attacks like these. And they're annoying features regardless.

the thing about “never attribute to malice what can be easily explained by incompetence” is that it’s rat-fuckable

when there is functionally no difference between the two, engaging with someone as if they’re incompetent means accepting their frame, that what they’re ultimately trying to accomplish isn’t *bad*, they’re just going about it in a way with bad side-effects, and people use in bad-faith our good-faith willingness to treat them as incompetent to push their agendas

engaging with someone as if they’re malicious, on the other hand, means rejecting the harmful frame, recasting the argument in terms of “why are you trying to do this bad thing?”, and not quibbling about the details of why the thing is bad

these age-verification laws whose implementations are a form of category error is a good example; if you engage with a proponent of them with “well here’s why your implementation is bad” you’re tacitly approving the larger idea that surveliance is good, and you just disagree with the techniques; bad-faith actors use this

If instead you come back with “why are you trying to surveil everyone’s computer use? Why are you laying the groundwork to prevent people from using their own computers?”, you re-cast the frame. Sure, there are probably incompetent people who don’t realize the results of what they’re going to do, but casting the larger idea into question AND KEEPING IT IN QUESTION is the only effective path I’ve found to debating people on things like this

so, instead:

don’t ascribe to incompetence something that is functionally malicious

So, you want to start a tech co-op?

Monthly videocall in the internationalist union hall for folks like you, coming up in 20 minutes!

📆 Last Sunday of each month 17:00 UTC
📃 meeting notes: https://pad.data.coop/56HJK2TYTvSeFx-ye48OXw?view#
🤙 jitsi link: https://meet.jit.si/techcoops

#tech #coop #cooperatives #cooperativa #democracy #freedom #solidarity #anarchism

As proof that LLM generated code is easy to audit & secure, I notice that anytime I let my smart friends look at it, they immediately spot an exploitable security vulnerability. Try getting feedback that fast on old style human code.

Latest comic: The post-human economy

#labor #work #economy #workers #inequality #comic #cartoon

RE: https://flipboard.com/@scimag/news-from-science-dgvs7ea7z/-/a-w1XCnGh5SY2WxkYQRcpUOg%3Aa%3A2427788627-%2F0

This is a much bigger deal than anything that happened at that hotel tonight.

518,428 children have died preventable deaths in the year since Musk dismantled USAID, saying, “Time for it to die.”

how about charging musk with 518,428 charges of negligent homicide?

#NegligentHomicide #USAID #politics https://www.motherjones.com/politics/2026/04/geek-tragedy-altman-musk-bezos-zuckerberg/

I joined #Fediverse after 2025 inauguration. So far I see the future is bright here at Fediverse

The clean energy kind of bright

Never going to be as bright as tech-bro-business kind of bright like #Facebook, #Instagram, #Twitter (never going to call anything else), or #Bluesky, but that is okay with me

I choose clean and friendly energy kind of bright over fossil fuel polluted kind of bright.

And that is how I explain to people why I only have #Mastodon and #Pixelfed accounts when they asked

For now, have a good night!

CARTOONIST DEFENDERS in the Sussex area, UK are encouraged to attend Monday evening’s Cartoon County event in Brighton where our Executive Director Terry Anderson will be speaking along with John Curtis – founder of Africartoons, and the newest member of our board of directors.

https://www.cartooncounty.com/nextguest

Hi #socialcoop and the greater fediverse 👋 nice to meet you.
We're a democratic #techcoop from Stuttgart, Germany and we're specialized in web software development, website design and #opensource service #hosting. We support organizations working for the common good and aim to build lasting connections in solidarity with them. Let's talk about open source, technology, democratic project and organization governance and solidarity economy.
For more check our website https://coquest.coop
#introduction #newhere #neuhier

Why let A.I. write your manifesto when I did it for free? https://www.dailykos.com/stories/2026/4/24/800026575/cartoon/tech-bro-manifesto/

A few years ago, after finishing 20 years as a sysadmin, I compiled a list of rules that were generally true.

1. You can't automate a process that doesn't already exist

2. Good engineering makes for easy administration

3. The easiest thing for us to do is exactly what we're told. Inventing what needs to be done takes longer.

4. It's just as easy to do something right as it is to do it wrong (or avoid doing it)

5. Everything you know will be suboptimal in three years, replaced in five, and obsolete in ten

6. Someone who is unwilling to learn is no different from someone unable to learn

7. It costs more than you think it will. It will also take longer, and do less

8. Extent always exceeds intent

9. The only safe connection is a disconnection

10. Your intellectual work is a part of the total cost of ownership

Not enough people are aware that there is an old Scottish song by the title "Cock Up Your Beaver."

That is all.

https://en.wikisource.org/wiki/The_Book_of_Scottish_Song/Cock_up_your_beaver

Well-documented list of free software projects and their use of genAI:
https://codeberg.org/small-hack/open-slopware

It’s already a long list that shows what looks like uncritical adoption, both by high-profile projects (systemd, VLC, etc.) and by niche projects (GNU Mach is a prime example).

Help us secure our warehouse for the future of left ideas. Every piece of merch sold gets us closer to our goal.
Shop new merch here: https://pmpress.org/index.php?l=product_list&c=255
Donate to our GoFundMe here: https://www.gofundme.com/f/book-by-book-brick-by-brick-pm-press-celebrating-20-years

Libadwaita demo app on Android, new update for Goblint, a linter for C GObject codebases, Java/Kotlin library for interacting with XDG Desktop Portal and much more in This Week in GNOME!

#246 Offline Dictionaries
https://thisweek.gnome.org/posts/2026/04/twig-246/

#ThisWeekInGNOME #GNOME #TWIG