picklefactory

Marriage A-la-Mode, Plate V: The Death of the Earl, by Simon François Ravenet, after William Hogarth, 1745, 📸 by @evanvucci

I could not do the @standardebooks.org stuff I need to do without it, thank you 🙏

Diana Wynne Jones wrote four or five "young wizard at boarding school" YA novels, each in a different but related continuity, and each of them is better than any of the HP novels in my opinion

... I did not know Smolitaire was a catalog game, I've been playing a sideloaded version this whole time??? must fix this at once

nightmare blunt rotation where each of them has a contrarian take on the blunt construction, strain, and direction to pass

genuinely less functional than a one-dimensional city in the desert

feel like I was just yelling about this a couple days ago 🫠

considering I definitely like the matrix analogy more than the original blogpost one 🙃

🤷

I dunno, by analogy the ratio of accounts might be generally correct. how many accounts are not hosted on bluesky PDS infra? is it up to 0.05% yet?

have to make sure MAGAs keep their hand down on the stovetop for the next three years with the rest of us.

both books are excellent in a way that stands out, and besides being incredibly different from each other, they are also excellent in different ways!

burns i mean, curse you brain for misdirecting my posts

what a fine goggins! and jere davis looks like he should be in a 1940s newspaper strip as The Criminal Element and that is a compliment

stridently, endlessly wrong about something? *this* guy??

making people uncomfortable in the short term about their magic decision-making technology seems more and more to me like an excellent idea we are too far behind on as a society

the oppositional dynamic of redundancy vs. efficiency also has a part to play here. what will this do to delegation and decision-making in the long term?

I am not optimistic about the way we will collectively answer this question

does one need that kind of variety for decision making? not in the short term, or not for any individual decision perhaps. but what about the long term? what about the next million decisions?

...or "You want candy sprinkles on your fried eggs this time", then I might think they produce variety of the same "type".

What do we get now? Perhaps I don't use one enough to know, but it looks to me like Always Yes And.

Anyway, LLMs as variety production machines are constrained by their data, training, weighting, and runtime environment, no? if someone trains and weights one so that sometimes we get "No and fuck you"...

I had a longer answer to this when I was hiking yesterday, but I have a shorter one today. Mainly I think this applies to language generation and not quite in the same way to code generation (I have some parallel thoughts but there are important distinctions).

not sure I can buy the idea that a human and an LLM will generate the same type of variety

Turn off BlueSky verification badges by navigating the web app to Settings → Moderation → Verification Settings and checking the "Hide verification badges" option box.

"we knew we would get moderation wrong" still reads to me as a misunderstanding of cause and effect

having been on the internet forever, it's such an ancient habit to be running services for myself–I "forgot" that people don't have that, nothing to give their data to the world for them, and the PDS provides this to them in a brilliantly portable way

it's good stuff, getting it published at one of my domains was a snap. I was using it to write HOWTOs for my dad the other day. but I also like the demonstration of just blasting a static site out from PDS to web

why should it help them affordably, when instead we can have one or two companies work to capture the whole thing and then extract rent from customers who now must navigate a shrunken and disrupted market?

no idea what you are talking about. my year-long strategy of muting anyone who might show me something like that has seemingly paid off!

has a new type of guy been discovered? I hope so

I was here around that time and I remember it. I guess I would say that if you onboard 20m users to your infrastructure and hire a moderation team, you are now planning to do moderation at scale regardless of whether or not you explicitly said so or wrote it down in a document