Cassandra has had an adventure

LLMs are fascist technology. Note that I didn't say "cloud-hosted LLMs" or "LLMs that aren't local-only" or some other qualifier. I did not, in fact, stutter.

I have approximately one spoon today. Do I spend it taking photos and/or writing fiction? Do I spend it writing a gemlog post trying to beg people to stop doing this shit? Do I rewrite the work I did last week to develop a metadata cache for Calibre-style libraries, work that's now unusable because SQLAlchemy went slop? Do I move my photography business off Pixelfed? Notice how many of those possibilities are tasks that only need done because someone else was an asshole.

Every single day, I wake up and find someone else who once gave me a gift has now betrayed me. It is incredibly straining on mental health, to say the least, and it is also a giant DDoS on time and effort.

It is so incredibly bleak watching one OSS project after another betray every single principle, betray their users, betray *everything* to cuddle up to fascist technology.

It would be nice to have multiple spoons, but I am in a single-spoon regime because no one with the capacity to make the world better fucking took COVID seriously. There was, in the US at least, a bipartisan effort to ignore COVID and let it continue running rampant rather than spending a relatively tiny amount of money on sending people masks and funding HVAC upgrades. Again, an asshole or bunch of assholes made life more difficult for everyone.

That preface aside, here's what no one asked for but that I need to say something about. AI is a critical attack on trust, and we should treat it as such. AI is perfectly poised to inflame passions, create wedges, drive people apart, and break communities into more conquerable cliques. We have *ample* evidence of this.

My point and objective with making such a strong statement without any caveats and nuance, though, is to draw a line in the proverbial sand: LLMs are fascist. Deciding how we deal with that and respond to that is much more helpful, I think, than trying to debate to what exact extent LLMs are or are not fascist. Once we admit that they are a problem, we can have a useful discussion about how to address that problem.

RE: https://hachyderm.io/@inthehands/117326716308932923 It's only been an hour, but already I'm kind of obssessing over this toot. The original post is about OSS and LLMs, making an incredibly necessary point in that context, but I think it goes much deeper. There's a radical statement made by FOSS principles, namely that source code is meant to be read and understood. Auditability is pointless without legibility and transparency.

I am so very glad that Pixelfed reversed course and is now banning LLM commits entirely. https://mastodon.social/@pixelfed/117350705262301318

The pattern is by now well-established. Slopbros will cause problems on purpose. Someone overworked and underappreciated needs to solve that problem, and does so in a way that isn't literally perfect in every way. The rest of us are left wondering if they are also a saboteur or simply exhausted and trying to do what they can. Conflict ensues. This is the goal. Once conflict ensues, it's easy for the sloppers to lovebomb, it's easy for them to appeal to reason.

Oh, another thing I could spend my one singular fucking spoon on: finding a new Gemini browser because maybe *just maybe* there's a way to make a web that's small and undemanding enough to reach sustainable permanence. https://jawns.club/@skyfaller/117350026517002874

Linux, Python, GCC all have slopware commits now. Trying to create a fascism-free dependency chain now means making an entirely new OS and compiler toolchain. I fucking hate it.

Anyway, as the guy who opened up the issue proposing that KDE explicitly allow slop PRs quite validly points out, it's not really up to anyone but KDE contributors. That's fair. But what the same guy is a thousand percent communicating though, is that KDE is not for people like me. It's not for people who have something to lose to fascists. If KDE developers don't want to stand for KDE's users, there's no obligation for them to do so, but let's not mince words about it either.

You know what, let's open this up to the floor. If you're tired of complaining about AI, share some absolutely petty complaints! Please don't share hot takes or start flamewars, but let's just get some of the minor irritating shit off our chests for a moment, the kinds of things that just grate.

True story: I got those headphones with my five-year gift money from Microsoft instead of a custom etched crystal because I figured the RGB kitty ear headphones would piss off one particular transphobe (not even one of the transphobes *at* Microsoft), and that seemed more valuable than laser etching that I could do at home.

To take another example, the US is now effectively, if not explicitly, a fascist state. That does not mean that if you live in the US or otherwise interact with the US government, you are a fascist. It means that the US is an economic powerhouse (though it's working on changing that, that's gonna continue to be *fun*), and the world sometimes makes you deal with that in horrid ways.

Gee, why might a bunch of people who have a lot to lose to fascism get a bit heated when you ask if we should do a fascism?

Um, what else can I complain about that will take my mind off of AI? Uh, gods, I miss dedicated video players instead of having separate video players for Netflix, YouTube, HBO, Paramount, and whatever that all are half-assed in distinct ways and do inane shit like putting pause buttons over the actual video I'm trying to watch.

This is not academic to me. It's very analogous to the situation with Calibre — not identical, but analogous. As lead developer of Calibre, Kovid had zero obligation to his users. He was *and is* free to betray Calibre users with impunity. I don't hold his choice to do so in high regards, however, nor do I hold the guy opening that issue on KDE's AI policy in any kind of particular respect. I hope that the rest of the KDE crowd tells him to take a hike and stop making things worse. We'll see.

About a year ago, though, this AI fuckshit hit something I care deeply and passionately about: books. The lead developer of the e-book manager I depend so heavily on to enable reading and delighting in the written form decided to betray their own wonderful gift, and to undermine the artistry of the written word with AI. Now I'm diving back into open source again with a very singular purpose — to protect art from AI. I know why I'm doing this, and that helps more than I can say.

I've "used" LLMs in the sense of collecting information to criticize them, but also in the sense of "accidentally pressed a button when someone moved my cheese again" or "called a medical provider's billing department and was forced to interact with an LLM-backed chatbot" or "used DuckDuckGo and it generated an AI summary on my behalf without asking." (Yes, I know about DDG's noai flag. Sometimes I forget to use it.)

I don't know why so many people need to hear this, but you don't fight fascism with AI, and your OSS is meaningless if it betrays users as completely as do proprietary OSes.

But I do feel like I need to say something about the KDE / AI discourse, and for two reasons. First, because I said things before and a follow-up is needed. Second, because I aspire to maintaining a fork of software that went through a similar round of bullshit, but under very different circumstances. I do think I need to say something, and that it's not just the social media five-minutes-of-hate trick again.

INT. QA LABS — DAY A 10X DEVELOPER sits at a chair, examining a variety of special purpose libraries, languages, algorithms, and other tools created over the past fifty years. An EXPERIMENTER hands him a new task, shaped exactly like one of the tools spread out across the table. EXPERIMENTER Which hole does this one go in? EXPERIMENTER (con'd) That's right, the AI hole! From behind the one-way mirror, we hear the faint echo of a defeated scream.

Relatedly, I don't know why so many people need to internalize that just because there's a hole doesn't mean you need to make it bigger. Open source was already broken in a few important ways, but instead of fixing it, we're witnessing a generational shift towards active and enthusiastic sabotage.

The exception being their kitty ear RGB Bluetooth headphones, which have an absurdly long battery life and have held together quite well.

Perhaps in the spirit of @glyph@mastodon.social's post yesterday about having sharp reasons for decisions in open source, I will invite anyone currently volunteering in open source who does not have a sharp reason for doing so to take a moment, a day, a month! off and come up with a sharp reason. My point here being that if you don't know why you're doing a thing, no matter how positive that thing is, it's that much easier for your good will to be subverted towards ill. Clear reasoning is defensive.

"Why is Proton cross-marketing with Framework," she asked, fully knowing the answer but feigning ignorance out of exhaustion.

(As ever, I want to make sure that I am both thanking Kovid Goyal for his nearly immeasurable gift and also criticizing him for deciding to betray his own efforts. Both can be, and I posit are, true.)

@nivex@tenforward.social god *fucking* damn it.

I'll give my reasons, by way of putting my proverbial money where my proverbial mouth is: I worked in quantum computing for approximately twenty years, making damn sure that whatever public money put towards QC research was backed up by open source projects and not locked away from the same public that funded it. When I transitioned, I was discarded like so much garbage, despite transformational contributions to the field. So I left QC and tech more generally. (1/2)

Not that I ever got around to doing said etching of course, and the unetched crystal itself is sitting in a box somewhere, used as a display riser. lolz

@cthos@mastodon.cthos.dev @nivex@tenforward.social If only there was a resource for figuring out which projects have gone slop. I joke, because yeah I fucked up with how I contributed to open-slopware, but I do think there needs to be *something* of that form, if only just to inventory the damage that will eventually need repaired.

@discatte@digipres.club @atax1a@infosec.exchange @ireneista@adhd.irenes.space Software needs far more unions, both of the labor and discriminated kinds.

...and he's still going. As a connoisseur of posting through it, dude is hard core crashing out. It's a hell of a thing to crash out over, though: that some trans people may not want to vote for virulent transphobes, and that maybe we should demand better of an opposition party.

This seems really neat! https://pypi.org/project/forgejo-ntfy/

You. Just. Don't. Gotta. Hand. It. To. Them. They're a bunch of upwards failing fascist conmen, you don't have to take them at their words. To a one, they're the kinds of fuckers such that if they told you the sky was blue, you should instinctively reach for your umbrella, rain jacket, and flood insurance endorsements.

Somehow, a lot more folks seem to get that when Trump opens his mouth. Not enough, but more at least. But when these absolute tools speak, tech media collectively trips over their own shoelaces trying to rush forth and do a stenography. Like, come the fuck *on*, the difference between Musk, Altman, Amodei, and Trump is... um. Give me a minute.

He's *still* going, and I'm *still* getting second-hand exposure to his rancid, rancid takes because bsky.app URLs don't make it clear who is being linked.

I wish I could block Bluesky profiles without signing up for Bluesky PBLLC terms of service.

brb, making Chrome extension to replace every one of those names with "fascist failson."

If something is worth doing, it's worth doing on purpose.

@love@thistlenfern.org I do find brid.gy helpful for following the few trans journalists over there, but if it weren't for that, yeah, I'd block as well. I am continually upset at how much of trans-focused journalism plays out on Substack and Bluesky, two companies that very much so seem to want us dead.