Cassandra has had an adventure

Sometimes I write intimate eschatologies or words about technology and math. Sometimes I make things by burning them with light or squeezing them through a small, hot tube. Sometimes I push water with a stick while sitting in a tiny boat.

Pronouns: she/they
Joined Fediverse On: June 15, 2017
Website: https://www.xgranade.com/
Newsletter: https://newsletter.xgranade.com/

Public Key

npub1hzvzdrx8mnxqy9m4xupkce4krf564qac2yfpupw9w80r88z8zkxsw5lzy3

NIP-05 Address

xgranade_at_wandering.shop@momostr.pink

Profile Code

nprofile1qqst3xpx3nraenqzza6nwqmvv6mp56d2swu9zys7qhzhrh3nn3r3trgpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsd4xf9s

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1hzvzdrx8mnxqy9m4xupkce4krf564qac2yfpupw9w80r88z8zkxsw5lzy3

Show more details

Last Notes

2026-05-14 20:46:17 UTC

There's something truly dismal about watching foss go slop. Like, free software was supposed to be about ideals, a better vision for what computing could be. The rampant misogyny, transphobia, abelism, and racism already showed how limited that vision was, but now it seems like it doesn't exist at all.
Just... volunteer work for giant corporations. The same extractive vision of computing, but now coupled to a Reagan regime oppression of labor rights and environmental deregulation.

2026-05-06 17:48:33 UTC

- reply

I think my point still stands? The state of math education sucks, but that doesn't make the math less important. In particular, we shouldn't lose important semantics like that a series doesn't imply execution order (or indeed, execution at all).
It's fine as an analogy, so long as the limits of that analogy are kept in mind?

2026-03-20 05:31:56 UTC

A few thoughts on Astral / OpenAI, now that the emotions have sat for a bit.
First, let me start by noting that AI is an attack on open source, inherently, by necessity, and at a structural level. That argument is bigger than Astral, but the short version is that you cannot simultaneously expand the public commons and work towards it's enclosure; moreover, if the public commons do not stand for the public good, then it's not really a commons any more.

2026-03-19 04:36:59 UTC

Thoughts on my first Apple system:
- I don't remember having to give my phone number to Linux when I installed.
- I think I've seen three UI toolkits in the setup screen so far?
- Dear *gods* I have to hit "skip" or "not now" or "maybe later" a *lot*.

2026-03-16 03:13:04 UTC

- reply

"it would be funny if microsoft managed to get a clue"
yes, yes it would

2026-03-15 18:23:09 UTC

- reply

made the point that the Neo is an implicit promise from Apple that macOS will run just fine on 8 GB of memory for the next 8 years.
But I think it goes farther than that: Apple made a reference device for application developers. They've never been shy about enforcing requirements on developers, and this is an interesting positive side to that: developers now have a huge incentive to make applications that fit within modest memory limits.

2026-03-14 23:51:08 UTC

The problem I'm trying to solve, yet again:
- I'm trans, and I hate my fucking voice so much I get dysphoric when I hear it.
- I have a local-only no-cloud voice recorder that I would like to use for taking notes on the go.
- I have a device I made (kind of, it's in progress) for backing up things that look like USB mass storage devices to self-hosted storage.
- I would like to be able to plug my voice recorder into that device, and have notes come out, without having to hear my own voice.

2026-02-20 22:28:27 UTC

- reply

I'm not personally sure I'd go quite that far. Someone I respect said something I vehemently disagree with, it happens. I don't think it'd be reasonable of me to insist that my heroes, as it were, agree with me on even all the important things.
I'm only being as pointed with the above because what he said is *also* something a lot of far worse people have said, and now they have more ammo.

2026-02-20 22:08:55 UTC

No, opposing LLMs isn't "purity culture." I've seen this now from quite a few different people, and I disagree vehemently. It is good, actually, to have moral principles and hold to them, even when people with more money than you find said principles annoying.

2026-02-07 23:28:34 UTC

- reply

As an addendum, let me emphasize: if someone is making 10^{5.5} dollars per year, they're rich. My point is that economic disparity is so incredibly bad in the US that being "rich" doesn't mean one has anything meaningfully in common with "owner-class rich" in terms of political power *or* security with respect to having basic needs met. It is a mistake for someone who is rich to think that it is not in their best interest to show solidarity with other laborers.

2026-02-07 12:02:47 UTC

- reply

So yeah, a lot of folks in tech believe the lie, think of themselves as owner-class instead of labor-class. A lot of shitty, shitty politics follow from that core untruth. Not all, not being reductive here, but a lot.

2026-01-22 20:36:09 UTC

Maybe it's nothing. Maybe I should go back to my usual arguments about why AI is bad. Maybe it's a coincidence that several high-profile boosters have recently had very public mental health crises of a form that appears to be novel and related to their own AI usage. Maybe it's the case that AI *just* is an attack on labor founded on eugenicist thought that enriches fascists and requires astronomical amounts of energy.
But maybe we're seeing something new about what LLMs do to human brains.

2025-09-22 03:34:39 UTC

- reply

Is that even possible? The actual Charlie Kirk was a complete ghoul of a human being, despicable through and through, yet he was *popular* with a nontrivial group of people.
If you have a character that's supposed to be a warning about hatefluencers* like him, how do you make that character unlikable even to real person's fans?
(*Sorry for coming up with the term. But it's what he was.)

2025-09-22 03:32:10 UTC

Had an interesting conversation about how to represent villains in media, partly prompted by a fictional story that wound up accidentally being about Charlie Kirk, despite having come out well before the shooting.
But still, it's an interesting thing to reflect on sometimes: how do you make a character so abjectly terrible that no one will root for them?

2025-09-15 07:42:59 UTC

- reply

The fediverse has had its near-netsplit kinds of arguments, to be sure. For example, I still maintain that assuming good faith on the part of Facebook, of all companies, was and remains a bad idea.
For all its foibles and follies, though, this remains one of the very few parts of the internet where you can safely call a violent racist, abelist, misogynistic, transphobic nazi shitstain what he is without getting banned for it.
That's really important, and something wonderful to treasure.

2025-09-15 07:37:09 UTC

Maybe one of these days I'll be foolish enough to write a longer post about it, but *wow* did the protocol discourse last month ever leave me still more sour on ATProto than I was before — and that was even before Bsky-the-company went to bat for Kirk.
Obviously, you do you, but I definitely came away from that discourse with the understanding that work on ATProto is, until something substantial changes, concentrating power in corporate hands.

2025-09-11 22:21:30 UTC

This is Bluesky's moderation policy, in case anyone is wondering why I want nothing to do with Bluesky PBLLC, ATProto, the whole mess.
(source: sent by someone I know via a private Discord server, shared with their permission)
https://stockroom.wandering.shop/media_attachments/files/115/187/988/701/945/328/original/dad66c11ec341707.png

2025-09-01 20:04:51 UTC

- reply

Intel Inside logo meme, but "Intel" → "Dead", and printed on a sticker that can be slapped on the doors to abandoned malls or other privately-owned third places that corporations once saw profit in maintaining. Alternatively, on shuttered libraries and other public institutions that have been defunded under obscene austerity measures.

2025-08-25 18:33:55 UTC

Reading the Google blog post on their new Android lock down, and it sounds both far less bad and far worse in different respects?
• Only applies to devices with Play Protect enabled, which IIUC excludes Android forks like GrapheneOS?
• The only app review is linking apps to a government ID.
• Wait, government ID? That's fucking terrible!
• The only app review *for now*.

2025-08-18 23:18:08 UTC

- reply

Just fuck it. Fuck C, fuck C++, fuck that whole gatekeeping culture.

2025-08-18 23:17:01 UTC

- reply

Like, c'mon. We can put the toxic git-gud approach to language and toolchain development behind us. We don't need to haze people out of systems programming and embedded development.
Of course, I say "we," but it's not actually "we" any more. Because what I've learned is that you can never, ever be good enough for the C/C++ crowd. Ever.

2025-08-18 23:15:34 UTC

Of all things in fediverse culture, the one that will always get to me is that shitting on Rust is such a sure fire way of getting boosts. The tech gatekeeping on here is sadly very real.

2025-07-30 08:11:24 UTC

- reply

There's something truly toxic and fucked up about the idea that children aren't independent people whom we owe a responsibility to, but are effectively a kind of property that can be controlled and coerced.
That view, enshrined in the incredibly fucked up framing of "parents' rights," is the same view that runs through KOSA, the UK's new censorship regime, YouTube's new "AI" filter, and so many other things.

2025-07-30 08:07:58 UTC

- reply

It's obviously not the case that children only lost their Constitutional rights in the US starting in 2007, but the Supreme Court deciding precedent that kids are effectively subhuman when it comes to legal rights was a huge turning point.

2025-07-30 08:05:29 UTC

- reply

This might sound like a shitpost, but I promise you it's not. In 2007, the Supreme Court held 5–4 that children do not have First Amendment rights, in that a school is legally allowed to censor their speech *even when said children are not at school*.
https://en.wikipedia.org/wiki/Morse_v._Frederick

2025-07-24 19:09:08 UTC

I understand, but I don't understand, how approximately ten anti-feminist reactionary women can bring every major credit card company into line with their hatred of all things horny and/or queer.

2025-07-24 00:49:17 UTC

- reply

Seriously, though, now do you believe me about AI companies aligning with fascism?

2025-07-24 00:48:10 UTC

Cursed madlibs. Cursed fascist hellscape madlibs.
https://fed.brid.gy/r/https://bsky.app/profile/did:plc:hf7ezrajxadu7v3tzcyij424/post/3luoawz76v22a

2025-07-15 23:35:43 UTC

God damn it, why the fuck is Jupyter pulling this shit?
https://ipython.readthedocs.io/en/stable/config/details.html#llm-suggestions

2025-07-13 16:58:18 UTC

- reply

uncertainty being epistemic: you don't know the truth about something, but there is a definite answer that you could in principle learn
uncertainty being ontic: the actual underlying truth is itself uncertain.
Pretty much all uncertainty you will run into in your own personal life is of the first kind. There is an answer you could in principle know, but for whatever reason don't currently know.

2025-07-13 16:53:15 UTC

- reply

I am begging people to please find a second analogy for uncertainty.
(Setting aside that *the whole point* of Schrödinger's cat is that uncertainty in quantum measurements is ontic, not epistemic, a conclusion he found unpalatable and thus decided to mock.)

2025-07-13 16:50:05 UTC

No. Schrödinger's cat isn't a thought experiment that makes any goddamned sense. It was designed (by a serial pedophile, no less) to be confusing and ridiculous.
Just drop it. You do not need Schrödinger's cat to make sense of anything, not sexual assault (e.g. "Schrödinger's rapist"), not the genocide taking place in Gaza, not even quantum mechanics itself.
You will be better off just explaining what you mean than hanging onto a pedophile's shitpost.

2025-06-30 19:00:42 UTC

Oh, fuck, Zoom updated.

2025-06-12 18:08:54 UTC

I'll never cease to be amazed and horrified at people who are more upset about a burning dumpster or a few burning robotaxis than sweeping someone's entire livelihood into the trash, disappearing and kidnapping people of color off the streets, or straight up murdering Black people. Property damage is not the same as violence against living, breathing people, and folks need to get some real perspective. Like yesterday.

2025-05-19 18:03:19 UTC

Thesis: ChatGPT is destroying education, students keep using it to cheat on homework and exams.
Antithesis: ChatGPT has no understanding of facts or semantics.
Synthesis: Homework and exams don't measure understanding of facts or semantics, and can be fooled by plausible-sounding bullshit.

2025-04-10 22:44:15 UTC

- reply

There's like no contrast between UI elements now, it's all black on black with very thin borders and white dots on *everything*.
I don't hate flat designs, I even kind of like them, but you need a bit more delineation than this to make it work. Good grief.

2025-04-10 22:41:34 UTC

Why does Discord look entirely different again?

2025-03-25 06:51:46 UTC

If you run a program that's deterministic, then by definition, your experiences with that program will be typical of its expected behavior. There's some important nuance there about making sure that the input you provide is the input you mean to provide, hence reproducible builds, containerization, etc.
But still, that's by contrast with programs whose behavior is inherently stochastic. There, it's easy to fool yourself into believing the program "works" when you just got lucky.

2025-03-25 03:40:24 UTC

If you are an AI proponent, please consider: what would falsify your claim that LLMs are useful?
To be sure, the burden of proof is on those claiming a new technology is useful and can be applied ethically, but even just for a moment, consider what it would take to convince you *personally* that your claim is inconsistent with observed reality?
If there is no such thing, if you are always able to shift to a better future LLM to avoid criticism, then that might mean your claim isn't empirical.

2025-03-16 02:15:32 UTC

The English language can, at a rough cut, be divided into several groups of words:
• Those words I can spell correctly.
• Those words I can spell surprisingly correctly.
• Words I get about 50/50 on.
• Words I can never spell right, but can always get close enough for spellcheck.
• Words for which spellcheck completely gives up on me, and for which I need to find some Wikipedia page mentioning the word so that I can copy and paste it.

2025-03-04 17:28:22 UTC

- reply

If a voting system leads to outcomes in which a preference for killing marginalized people wins an election, I will argue that's a moral failing of that voting system. In particular, I'll argue that that remains true even if other legal frameworks prevent or mitigate that negative outcome.

2025-03-04 17:26:59 UTC

- reply

To clarify slightly, there are absolutely legal frameworks, presumably including the United States Constitution, that address this issue. My point was that the assumption used to justify voting systems *in particular* seems flawed, or at best incomplete.
That is, I don't want people who want me dead to get elected, then find that they can't act towards that goal because of the Bill of Rights — I want for them to not be elected in the first place.

2025-03-04 17:06:40 UTC

Because I'm [REDACTED], instead of doing what I need to be doing, I just went down an almost hour-long rabbithole on Wikipedia about voting systems.
It really strikes me that the idea that voting systems should accurately represent the opinions of the majority is just taken as sacrosanct.

2025-02-08 20:48:29 UTC

- reply

I could dissect why; perhaps the FOSS movement could have better prepared for this moment if the Right to Read had instead been the Right to Make Art, the Right to E-mail Your Friends, or the Right to Be Trans Online.
But ultimately, there's grief under it all. It's not praxis, it doesn't help us survive this moment, to reflect on what could have been. It is human, though, and I will posit that it's OK to grieve that. Perhaps I'm just being selfish because I do grieve it. Still.

2025-02-08 20:45:30 UTC

- reply

I grieve it because what we lost is what we so desperately need right now, in this moment. A fascist movement has successfully compromised huge swaths of human communication, including both social media and person-to-person communications — but we've lost so much of the resilience of the pre–Web 2.0 world.

2025-02-08 20:43:19 UTC

The internet has come so far and changed so much in my lifetime, but I also grieve almost every day for what we've lost along that journey.
Decentralized networks used to be the default, from IRC to Usenet. When they were centralized, adversarial interop allowed things like Trillian and Pidgin to flourish. So much more used to be client-side, so that you owned your data in a meaningful way. Perhaps most importantly, it used to be much easier to put your stuff on the web.

2025-02-07 05:09:32 UTC

No messaging app on a phone can be more secure than the phone itself, so why bother with Signal?
This is why threat modelling is so important: there are absolutely a *lot* of threats that Signal protects you against, even on a phone that leaks too much data to Apple and Google.
It is both true that Signal is very worth it, *and* that to get the most out of it across the most threat models, your phone itself should be as secured as you reasonably can get it.
https://kind.social/@EnfysBook/113958146467673520

2025-01-31 20:04:28 UTC

- reply

At the same time, it is still amusing and can, I argue, serve to inoculate people on the sidelines against some kinds of fascist rhetoric. Building morale and solidarity doesn't win arguments, and doesn't defeat fascists, but it's also not nothing. Pointing out absurdity is a tool, one that is wildly overestimated by liberals, but it is a tool nonetheless.
(2/2)

2025-01-31 20:02:49 UTC

Generalizing something I said in replies, but I have read many times and definitely understand why it's not effective to call out intellectual inconsistencies on the part of fascists.
As per Sartre, "they know that their remarks are frivolous, open to challenge. But they are amusing themselves, for it is their adversary who is obliged to use words responsibly, since he believes in words."
(1/2)

2025-01-23 02:22:54 UTC

- reply

It is, unfortunately, the best option I've found so far. I share your sighs, in any case.

2025-01-20 06:03:53 UTC

- reply

Like how do you go from "here's why microbenchmarks are important for detecting performance regressions" to "yolo, lets add an ouija board to every node in our server farm" in twelve months?

2025-01-20 06:02:48 UTC

This is absolutely the least important thing about the whole AI fad, but it is very surreal watching the same software engineers who harped on about unit vs integration vs regression testing, TDD, matrix builds in CI, error telemetry, and everything else they said helped make reliable and correct software all collectively going "ah, fuck it," and trusting everything to a broken game of Boggle.

2025-01-17 17:45:46 UTC

- reply

It's been a few days, with a lot of misunderstanding in replies. That's on me for not communicating my point effectively, so let me expand on and clarify.
Suppose you encounter "HTML isn't a real programming language" as an example of gatekeeping people out of tech. It's a real and really awful canard that techbros use, even now.
Very roughly, you can respond in one of two ways:
• Expand the definition of "programming language"
• Challenge the idea that programming is necessary to be in tech

2025-01-16 00:40:10 UTC

- reply

That also begs the question of *which* Portishead album? They're all very different from each other!

2025-01-16 00:37:56 UTC

The Stranger is always there with the news you somehow never really needed to know, but also kind of did.
Never change, Stranger. Never change.
https://stockroom.wandering.shop/media_attachments/files/113/835/236/213/120/381/original/d7482e673f715c91.png

2025-01-15 22:29:28 UTC

- reply

AI grifters have basically invented a new "fact," that LLMs can be "programmed," and built a whole epistemological chain to back that up. Where does that leave non-experts trying to critically evaluate AI claims?z

2025-01-15 22:27:16 UTC

- reply

It's another way that AI grifting is a bleak mirror of earlier "intelligent design" grifting. Academic publishing is a complete disaster, and at the same time, is held up by institutions like Wikipedia as a marker of scientific legitimacy — and to some degree, rightly so. After all, there's not a better system already out there that people outside of a field can use to evaluate knowledge within a field.
So we're left with a situation where someone can just piss in the knowledge pool.

2025-01-15 22:24:42 UTC

One more thought on the whole Wikipedia / "prompt engineering" thing: AI marketers have abused the absolute hell out of weaknesses in the academic publishing system in order to manufacture the idea that "prompt engineering" is real.
Namely, preprints exist due to extensive problems in peer review and for-profit publishing. It's become the standard in many nascent fields where for-profit publishing has less of a chokehold.
But then AI grifters capitalize on that to just say whatever they like.

2025-01-14 03:17:03 UTC

- reply

As with most of my hot takes, I believe the above about 63% seriously, and offer it by way of making a point more than as a literal truth.

2025-01-14 03:11:41 UTC

Hot take: "HTML is a programming language" is gatekeeping in effect, if not in intent.
That is, something need not be a programming language to be a the subject of highly useful and important technical skills — viewing all of computing through the lens of programming languages is inherently limiting.

2024-12-15 00:54:02 UTC

- reply

Absolutely. The nice thing about WebUSB on Chrome though, is that it also works well on mobile... that might be harder with a native+extension approach?

2024-09-16 20:21:07 UTC

I wonder how much people would be shocked to see how expensive "cost-cutting" measures in public institutions really are. Cutting down the spending you can see often has the effect of increasing the spending you *can't* see; a classic McNamara fallacy.

2024-08-22 00:08:35 UTC

DoorDash AI told me *without my consent* that an item I was looking at to help accommodate my allergy and eating disorder is "tailored for kids."
Fuck you into the sun.