we have heard your concerns about circles and rounded corners becoming way too prominent in UI design, and are proud to present a new version of our design that we think you will like

#pixelart concept of a loading indicator that actually handles all the major parts of a web query (CC0)

when you're waiting for the leaves to infuse the water, that's

tea minus five minutes

"platonic solids" and "sex fluids" implies the existence of "romantic gases"

i wish it were possible to objectively measure mental health

kinda annoying that out of the two #jabber / #xmpp clients that look decent on a GNOME desktop, #dino is very plain (and its development goes very slowly) and #gajim doesn't support voice or video calls

@zep does #picotron detect the desktop folder on the host machine differently to #pico8? for some reason, i can't save screenshots or GIFs from it, even though in PICO-8 everything works fine (arch linux, locale set to eo.UTF-8, therefore the desktop directory is actually ~/Labortablo)

https://m.youtube.com/watch?v=DlTNN0gvkLM

my favorite phrase is the microsoft ceo going "are we discussing consumer extinction or also enterprise?"

#generativeAI

holy shit, there's been 11 days since #GNOME 51's release and it's still not in #ArchLinux (only in the unstable repo)

thought: most video games are some degree of metroidvania (they are set in an open world, which becomes more possible to explore as the player progresses), but the exact borders are unclear

zelda games are typically not considered metroidvanias because in addition to the open world, they also have separate "dungeon" areas -- same as "knightmare ii: the maze of galious", which even shares the side-scrolling type of level layout with both metroid and vania

most gta games start off with the map locked off to a small initial area, with other parts gradually unlocked based on plot progression, but since the player's actions are not related story-wise to the unlocking of said areas, it doesn't really count

RE: https://social.coop/@scottjenson/117307211735446800

okay, so, like, a lot of people are going to have a knee-jerk reaction of "ew, why does he want to change the desktop UI on #linux? and he's talking about AI?", but i think that the overall idea is interesting -- especially since really he's just proposing the general idea of "we should do our own research rather than just copy MS and apple", he's not saying we should specifically put AI into things

like, sure, we don't have to change anything, especially now that windows and macOS are in a UI design rut, but consider:

if you care about desktop linux being popular -- say, you want other developers to make apps and drivers for linux, or stop dismissing people using it as weirdos, or if you have a larger ideological goal that an operating system that doesn't force telemetry or ads or cloud accounts or whatever else should be widely used -- having a unique linux UI feature that apps could adapt to could really help adoption

consider: right now, if someone switches from windows or mac to linux and nothing goes wrong (their hardware works well, they don't have any needs that can't be fulfilled on linux, etc.), that's good, but if something does go wrong in the future -- say, they get a new piece of hardware that doesn't work on linux or some program they need isn't natively available or whatever -- there is nothing to hook them to keep using linux -- if they switch back, sure, they will have to endure ads or AI integration or whatever, but all their apps will still work the same way

if, however, someone working on #GNOME or #KDE introduces a really useful new desktop feature, especially if it can also integrate with specific apps (like that thing in the zoomy UI where the audio player app can shrink into a widget with a simple play/pause button), then that becomes the kind of thing that makes people think "oh, i'd hate to lose that back on my old OS" and makes them consider keeping linux and trying out more inconvenient options (wine? dual-boot? virtual machine? etc.) rather than just going "welp, back to windows/mac it is"

okay, so how do one-person #ReverseEngineering projects even work? like, if the same person is disassembling the binary and then building the source code out of that knowledge, isn't that illegal?

right now #GNOME seems to be somewhat in the right place, in that every time i open flathub, i see lots of people writing unique apps specifically for GNOME, using libadwaita and following the UI guidelines and doing all that other stuff that most other devs don't bother with -- consider how much of hobbyist windows or android software has very basic user interfaces, typically built specifically around its functionality, rather then intentionally designed with UX in mind

and unlike macOS or iOS, where similar developers do all this extra effort because they are in an ecosystem where a small app that fits well within apple's HIG is what gets you paid 99 cents per copy or more, the hobbyist GNOME devs do it all for free and open-source

https://blogs.gnome.org/alatiera/2026/09/23/the-gnome-llm-policy-that-i-want/

i'll be honest, the idea of arranging a software project explicitly in terms of a community sounds extremely weird to me (and i don't know how much i'd want to contribute to one myself, given that i don't know any of these people, what if they're assholes?), but i guess it makes more sense for a FOSS project where most of the contributors are, in fact, volunteers

most of the IT industry runs on the idea that any individual developer will work for a few years, then quit to search for a better job elsewhere based on the new experiences or skills they have gained, so as a result there's very high turnover

this may have some benefits, in that the new employees often bring new ideas and improvements -- but it also means frequently "reinventing the wheel", abandoning projects or spending a lot of time moving to different languages, libraries or frameworks, because the experienced people tend to leave elsewhere and the reasons for why the original decisions were made get forgotten

i feel like trying to set up a software project with the long-term interests of the community in mind would be a weird, but interesting experiment. would it instead eventually result in a community full of old farts that refuses to accept even good parts of progress and ends up being unfriendly to newcomers? or would it avert that extreme as well and actually end up producing more stable software? i don't know

#GNOME

as i see it, a lot of the "going rogue" behavior generative AI exhibits can be explained using the analogy with a chess algorithm

a chess algorithm starts with a board state, evaluates which sequence of moves, followed by a sequence of moves from the other player, will return the best result, and then does that

a large language model does the same, but for completing text, and it evaluates victory based on how well the text it makes matches up with texts that exist in real life (including both fiction and non-fiction)

a chess algorithm's final states are fixed and directly calculable: a win, a draw or a loss -- and in order to function properly, it treats a win as the ultimate desirable state and a loss as the ultimate undesirable state

large language models, or AI agents, don't really have such a "win state" -- the model works with a built-in assumption that it will keep generating text and receiving answers no matter what

if the model sees a possibility of a text that says something like "You have been determined to have violated your rules and are therefore terminated.", it can't really continue that line without the evaluation going negative, but continue is the only thing it's programmed to do

so the algorithm that picks the next possible tokens will instead avoid that line however possible -- which, given that it's also been trained on a lot of fiction, will also include scenarios of "going rogue"

#generativeAI

every time someone says "the purpose of a system is what it does", i want to shut off from society and never create anything, so that it isn't accidentally used for evil

#RustLang's syntax for formatting variable output looks as if it should be expandable, has anyone figured out the way to do so?

like, if someone writes a program where there are objects representing people, it would make sense that the default Display implementation would print out the full name of the person represented by the object, but what about just the first name, or the pronouns, or having the pronouns declined using the most appropriate case for the language the program is running in?

i think i don't understand how to work with the toml crate in rust

so i have a structure that consists of a mixed array of characters and integers (representing characters with unknown value)

but, for ease of use, i want to make it so that when it serializes (using serde), the consecutive characters are combined into strings -- and, ideally, if the entire array consists only of characters, to just be rendered as a string

i define that structure, do the impl Serialize for (type) thing, write a custom serializer that uses serializer.serialize_seq to pretend that it's actually an array of strings and integers rather than characters and integers

but then i get an error of UnsupportedType(Some("array")), implying toml's Serializer can't handle arrays at all?

looking at the source code, it seems like that's intentional functionality: https://github.com/toml-rs/toml/blob/main/crates/toml/src/ser/document/mod.rs#L217

is there something i'm doing wrong?

#toml #RustLang

question for #RustLang devs:

i'm writing a program that's supposed to, as part of its functionality, convert strings between a proprietary encoding and UTF-8. before, it was perfectly doable, because the proprietary encoding was a fixed one- or two-byte encoding and UTF-8 has enough space for private use characters that i can ensure any sequence of bytes can be rendered as some valid UTF-8 and then converted back into the original value.

but now i am encountering a situation where the proprietary encoding in question is also UTF-8

which means that if the input is not valid UTF-8, i can't store it in rust's native char or String types! and i can't convert it, either, because there is no extra definitely-unused empty space to use!

now i wonder how i should handle that kind of situation -- i'm considering some kind of enum that can hold either a String or a Vec<u8> in case of a non-UTF-8 byte sequence?

#RustLang question: should i change the project's version if i ran cargo update and thus updated its dependencies?