Nefalšovaná kočka u korýtka. Jen pro silné povahy!
Sensitive content
Na veterinu s ní šel #autobusak, poněvadž já v práci. Doktor prohlásil, že je zdravá a dal jí nějaký vitamíny a konzervu. Jako bysme jí nepodstrojovali dost. Neudělal jí sono a ani rentgen, a že jestli se do pátku nezlepší, máme přijít zase. Takže radši začala žrát. Musím zjistit, kolik nás ta zdravá kočka na veterině stála.
Tohle není příjemné poslouchat, ale měli byste si to pustit. Máte na to spoustu času, nemá to časové omezení jako jiné (příjemné) pořady. mujrozhlas.cz/rapi/view/serial…
Koho mučí v ruském zajateckém táboře? Kam míří ruské údery? Jak zní válka a jak krajina předků? Kam a kudy utéct? Jak si navzájem pomáhat? A jak truchlit, přežít, nebo žít dál? V Radioknize uslyšíte exkluzivně sedm povídek, reportáží i esejů nejvýzna…
If your corporation is a monopoly (or part of a duopoly, or a triopoly, or whatever), you can fire all your best staff, let your services degrade to hell, and you’ll still keep all your customers (and all their money) because there’s nowhere else to go.
I don’t really care that all the big cloud providers are corporations beholden to the USA. I care that there’s 3–4 of them. It wouldn’t matter where they were based if there were 100.
libghostty can now encode terminal contents as HTML. On the left is an example program that reads stdin, loads it in a terminal, and encodes to HTML. On the right, an example of `fastfetch` encoded to the browser. Coming soon to Ghostty clipboard ops automatically.
I've long held the opinion that the proper way to expose TUI semantics (which nobody is doing at the moment, as there's no standard for it) would be to rely on HTML in some way.
We already have perfectly good screen readers that can work with complex objects (think tables with column headers spanning multiple columns) on the web, why re-invent the wheel?
Aria has everything we might possibly need, menus, progress bars, tables, forms (for 3270 / 5250 terminal emulators), it's all just there, well specified, well supported and ready to use.
@miki Thats a really good idea and angle for this... actually. Right now we're just emitting some very bare bones inline styling and very little structure. It'd be fairly trivial to augment the HTML formatter to emit aria tags and some more things to make it usable by a screen reader perhaps: github.com/ghostty-org/ghostty…
Consuming this would be another matter entirely, you'd have to write some Electron / native webview / browser-based abomination, communicating with some native process (with enough privileges to actually run terminal commands) over websockets. You can't "just" expose raw HTML to screen readers, you need some kind of web engine that converts it to the accessibility APIs that screen readers understand natively, and those varry per platform.
You could make Ghostty itself pretend to be a webview (which AFAIK would be the only way to communicate a complicated document representation like this), but that would either rely on the platform's native web view somehow, or re-implement those APIs.
Perhaps @matt has some input on this, he knows a lot more about this than I ever will.
@miki @matt My stance just purely out of pragmatism of being a hobby project has been to push all a11y work into one platform and that has been macOS. So I think leaning on WebKit and bridging that way via the native APIs could work fine.
@miki Does ghostty embed a browser engine yet? If not, I wouldn't do so purely for accessibility. A browser engine is a big, heavyweight component, and I'm *sure* that many ghostty users would be disappointed if a browser engine were brought into their presumably currently lightweight terminal.
@matt @miki It doesn't, but its a "standard" library for macOS apps so its dynamically linked and doesn't add to the binary size. If it were only used for unobtrusive reasons like a11y I don't think anyone would notice or complain.
Of course, you're free to do as you wish within your hobby project, but my professional opinion is that retrofitting a web view onto your terminal wouldn't even be the easiest way to implement accessibility on a single platform. @miki's idea of using HTML to make terminal-based TUIs more accessible is just hypothetical at this point, and I'm not sure I agree with it.
@miki To make the terminal itself accessible in the near term, you might look at AccessKit (accesskit.dev/), which I started. Yes, we still need to write more documentation. It's in Rust, but it does have a C API now, and the static library for that C API on macOS is only a few hundred kilobytes. It should be pretty easy to add that to your existing SurfaceView. (I've looked at ghostty just enough to find that class name.)
@miki There is in fact already at least one Zig GUI toolkit using AccessKit: github.com/david-vanderson/dvu… Though I don't know if you'd be using the AccessKit API from Zig code, Swift code, or some of each. It looks like it all gets statically linked together, so using AccessKit from both languages, e.g. cross-platform Zig code and macOS-specific Swift code, should be fine.
Not to mention that you'd first have to establish a spec for apps to expose the needed information in the first place. As it stands, terminal emulators don't "know" that you have midnight commander open, that it has two panes open, that foo.txt is selected in the left pane and bar.txt is selected in the right pane, that both should be treated as lists, and that the couple lines at the bottom are a status area. There's no standard to even expose that info in terminal land.
@miki Well yeah that’s a different category of the problem entirely and I’m unsure we’re anywhere near taking that on unfortunately. Making the raw text with command boundaries accesssible is the more short term play.
Mac OS has a significant problem with queuing. WHile Windows screen readers have a queue of utterances to be spoken, Mac OS can only support one utterance at a time. New speech output always clobbers what's currently in the buffer.
This is highly undesirable for terminals. Imagine a quickly scrolling compilation log, where all lines are prefixed with a timestamp. Instead of being able to periodically flush the queue and listen to a few lines to track progress, you'll constantly hear the word "two" (as in 2025-10-30). In other words, before a line can be read, it will be cut off by the next one.
This is the behavior of terminal.app (which is why it is unusable without muting VO and using a separate terminal screen reader, but that is an ugly hack that carries other issues with it.)
It's not an insurmountable problem by any means, you can get around it by interacting with Mac OS's speech APIs directly (and those APIs do support queuing, it's just VO that doesn't), it's just something to be aware of and that most terminal emulators get wrong, making them almost unusable on Mac accessibility-wise.
For an example of the problem in action, see this page with VO on (cmd+f5 or cmd+side button pressed three times on old touchbar Macs) gist.githack.com/mikolysz/9878…
Máš starší notebook nebo počítač a chceš zkusit žít s Linuxem místo Windows?
Chceš zjistit, jaké jsou otevřené alternativy ke komerčním cloudovým službám a běžným programům?
Tak pojď na náš celodenní workshop v sobotu 8.11. od 10:00 u náměstí Jiřího z Poděbrad v Praze, nainstaluj si s naší pomocí Linux a nauč se jeho základní použití pro běžné činnosti!
On this day in 1995 (thirty years ago), I experienced my first school fire, when someone threw a lit match into an empty soap Dispenser in an upstairs girl's bathroom, which was right down-wind from the library, which smelled bad for weeks afterward.
Disney and YouTube settle their legal dispute over YouTube's hiring in May of former Disney executive Justin Connolly to be global head of media and sports (Dominic Patten/Deadline)
I took two spectacular photos, one of Simba and one of Patches. Got errors when I posted, so I'm going to try again. The first photo is that of Simba, a gray and black tabby The second accessory is Patches, a calico with patches in her fur. Both cute, both great accessories worthy of bobble heads, air tag holders, or beer can holders??? Who knows. Too bad both of the photos continue to failed to send this message. If you guys can figure out how I can send these using Mona without them absolutely failing because I took it on an iPhone camera or something, let me know, and I’d be happy to show you how awesome these accessories are.
See, ambient authority is the natural and ecosystem-mindful way to write software. You take what you need from where you are. Ocaps are the capitalist, hyper-individualist survivalist model where you must carry everything you need all the time, and cannot procure it yourself - you need to be handed it.
MiniMax M2 is the new "most intelligent" open weights model (according to Artificial Analysis) - the MIT licensed weights are just 230GB and it appears comparable to Sonnet 4, while priced closer to Gemini 2.5 Flash. Notes here, including a new LLM plugin: simonwillison.net/2025/Oct/29/…
MiniMax M2 was released on Monday 27th October by MiniMax, a Chinese AI lab founded in December 2021. It's a very promising model. Their self-reported benchmark scores show it as …
Corporate America is so toxic and such a scarring ever lasting experience that even places that try to do better and have a transparent, open, and collaborative approach are seeing as leadership against developers, with deep trust issues, and an ingrained instinctual approach of trying to read between the lines, even when nothing is hidden. It's sad...so sad.
Cue the "you're a Director so you must make millions so go die in a fire" kind of replies
In 2013, when #Montreal was deep into its corruption self-investigation, we almost ran out of asphalt.
We'd made a new rule that anyone named during the Charbonneau commission's corruption inquiry was disqualified from being a vendor/supplier for the city.
The hitch was that all of the asphalt suppliers were named. Yes, *all* of them.
So we had to decide “run out of pothole-patching materials" or “continue doing business with corrupt suppliers”.
Twelve year old spoiler: #Montreal used the corrupt-suspect companies anyway. We had potholes to fill.
To me, this feels not so different from keeping my Facebook account for the beer groups, our business email on Google because of the spreadsheets/integrations, and even how @pluralistic talks about why he maintains his X/Twitter account.
It sucks, but there aren't alternatives that offer all of the same benefits (and I include “network of people" in this calculation).
@hub AFAIK the city didn't have an asphalt plant at the time. The linked article (in the reply) said that council voted to use the declared-corrupt suppliers.
Maybe there's an evolution I don't know about, though. 🤷♂️
A parallel universe where the concept of a "key board", be it a piano or a typewriter keyboard, never got popular. Most people are using fretted lute / guitar input with their computers; people who mastered chords can type faster than those who cannot. Some countries prefer violin/shamisen style of input, and thus in the 40 years of computerisation their spoken languages have changed to incorporate legatos and pitch changes. The biggest coder flame war of all times is not vim vs emacs but plectrum vs bow. A few historians uncover a long-forgotten prototype of a computer controlled by percussion.
It’s telling that Grokipedia, which is born of hatred and bile, is utterly dependent on Wikipedia as its source despite its contempt for it. Can hate sustain a collective project? Apparently not; it must syphon from others.
The days of the troubled Phoenix pay system appear to be numbered, as the federal government moves forward with implementing the new Dayforce system for human resources and payroll tasks.
Kleiner Service-Toot für alle die den Browser #Vivaldi unter Android nutzen: Laut einer Freundin ist dort offenbar kürzlich die Option verschwunden, mit der sich ein dunkles Theme für alle Websites erzwingen ließ. Die Einstellung ist aber noch in den Experimenten zu finden.
In die Adresszeile vivaldi:flags eingeben, und dort "Auto Dark Mode for Web Contents" auf enabled setzen. Browser neu starten, und schon ists wieder überall duster. 😀
Sandboxed Google Play compatibility layer: temporarily limit system service override infrastructure added for GmsFontProvider shim to Pixel Camera to work around certain banking apps detecting it as tampering (we can change the approach to avoid this to enable it for all apps using the Google Play client libraries again, especially since we want to expand it to improve app compatibility without Play services installed)
kernel (6.1): update to latest GKI LTS branch revision
kernel (6.6): update to latest GKI LTS branch revision including update to 6.6.112
kernel (6.12): update to latest GKI LTS branch revision including update to 6.12.53
All of the Android 16 security patches from the current November 2025, December 2025 and January 2026 Android Security Bulletins are included in the 2025102801 security preview release. List of additional fixed CVEs:
2025102801 provides at least the full 2025-11-01 patch level and the Android 2025-11-05 patch level (Pixel Update Bulletin could have fixes we don't get early) but will remain marked as providing 2025-10-05.
For detailed information on security preview releases, see our post about it.
Well! I've definitely seen my share of "idiosyncratic" C styles omitting braces and exploiting corner cases in 30+ years of having to read other peoples' C code, but this is a new one.
I had not thought about what would happen if you left the braces off a switch. Because, of course, doing so would be utterly unhinged.
I give you: the world's most obtuse version of "if (x == A || x == b)", indistinguishable from an editing mistake.
This whole idea that "AI" is a helpful tool for autistic people, because it lets you write more "professionally" or whatever, is incredibly toxic and frankly demeaning. Insulting, even. It buys into the whole concept of "one correct way" to express yourself in society (the neurotypical way) and that being your authentic self is wrong. Having ways of expressing your thoughts that don't conform to that dogma is something you should repress. Hide. Feel ashamed of.
I don't know why some people seem to take it as a given that you can't educate or persuade people, but I think it's a problem.
"Surely you can't expect [supposed type of person] to understand [idea]."
Well, not if nobody tells them.
I know we are tired of dealing with willful, malicious ignorance, but don't deny the reality that some people are able to learn new things & change their beliefs or behavior.
How do I know? That's me. That's my parents. That's my ex-evangelical friends.
That's exactly my objective: identify people who are open to having their perspective expanded & then help them see something new.
Often the idea seems to be that we must compromise because people can't be expected to come around on something. So my question is "who says they can't come around on that?" It's defeatism masquerading as pragmatism.
As someone who was raised in what was effectively a cult, who got out of it, who helped other family members get out of it, & who has continued to learn & change my mind & grow & who has watched the people I love do the same?
I fucking know that people can change. Drastically. Not everyone. Some people have decided to never change, & there's not much to do about that, but a whole lot of people are desperately eager for the truth, but don't know it yet.
The other day we celebrated everything curl turning 5 years old, and not too long after that I got myself this printed copy of the Chinese translation in my hands! This version of the book is available for sale on Amazon and the translation was done …
I used the LINE messenger for the very first time the other day. Downloaded from the official source, my goal was simple: a single, private chat with one person, which LINE claims is protected by End-to-End Encryption (E2EE).
Our "conversation" was minimal – mostly just test messages. One single contact.
A few days later, I log in to find my account has been temporarily (for how long? =)) blocked.
This raises a cognitive dissonance for me:
Promise vs. Reality: We're told E2EE means no one, not even the company, can read our chats. If that's true, on what basis was I blocked? I have to assume it's related to something in the chat, because I literally did nothing else in the app. But if it is E2EE, how could they know? Does this mean E2EE is just a marketing buzzword and monitoring is happening anyway? I'd genuinely like to know what the real factors are.
Absolute Opacity: This is the real issue. I received zero explanation. No email, no warning, not even a vague hint at which policy I "violated." And look, I'll be the first to admit I didn't read the 100-page Terms of Service – who does? But that's not the point. Even if I did technically violate some obscure rule, the core problem is the total lack of transparency. I was left with no idea, not even a guess, as to what happened. This opaque, black-box process is the real problem.
This situation is deeply concerning. What if I lived in a country like Japan, where LINE is the default, essential messenger? I'd just be cut off from my digital life without cause or appeal. And if this happens on LINE, what stops WhatsApp from doing the same? (And let's not even talk about Telegram, which is 100% cringe and a lost cause for privacy anyway).
My takeaway: To be honest, I went into this as an experiment, and this incident 100% confirmed my expectations.
This isn't just a LINE problem. We see it constantly from Big Tech like Meta and Google. They ban users, often with no explanation, because they have the full legal right to do so. We all agreed to this when we blindly clicked "accept" on their Terms of Service.
This is exactly why my advice is this: you must factor in this risk with all commercial messengers. When you use any private, centralized platform, you have to accept the fact that you can be denied service at any time, for any reason, and they don't even have to tell you why. That is the price of admission we all paid.
The promise of a "private chat" apparently doesn't include the guarantee of access to the platform itself.
No, E2E apps still block people without reading their chats. There are a lot of other signals they take into account, notably metadata, the OS you're using, how suspicious your messaging patterns are, your IP address, the phone number / other personal data you registered with, etc etc.
Many people in privacy circles unwittingly fall into these patterns.
I don't think you realize how necessary blocking actually is.
Imagine you have a country of 100 million people, each of which has a single account at your service (like most users have). Also imagine you have 1000 fraudsters. A fraudster, even when doing everything manually, with no computer programs to make them work faster (and that's not how fraudsters work), can easily make 25 accounts per day. That's ~9 million fraudulent accounts per year. This means that, after 10 years, *50%* of all your accounts are fraudulent accounts that are constantly sending spam.
We had a world of no blocking in the email of the early 2000s, and look where that got us.
@miki so, good example, thank you. I am not against blocking or even control (when agreed upon). This time, I just realized that what pisses me off the most about this is the opacity of the blocking. I also understand that, technically and economically, this is almost impossible, and that analyzing each user individually is not a viable solution. I will try to expand on my opinion later, which lies at the core of my protest: transparency and proper architecture in general, particularly when it comes to messengers.
This is a repository of apps to be used with your F-Droid client. Applications in this repository are official binaries built by the original application developers, taken from their resp. repositories (mostly Github, GitLab, Codeberg).
Ehh, this has always been a thing? Remote access always required a separate license addon, and features like Tandem Direct (or even running on professional versions of Windows) were pro-only for the vast majority of JAWS's life. It has only been in the last few years where licenses only differed by how you were legally allowed to use them.
Tak jsem o těchhle dvou expertech přeložila už šest sešitů. Napsaných v šedesátých a sedmdesátých letech, ale moji španělští kamarádi je hltali ještě v devadesátkách a od dětí okolo osmi let mívám pozitivní ohlasy i dneska.
Nevím, nakolik bych si Clevera a Smarta užívala jako čtenářka, ale na překlad je to velká zábava. Slovní hříčky s vizuálním doprovodem, parodie a dobové narážky, rýmovačky spojené s dějem. Pointy nesmí zaostávat za šílenou kadencí jednotlivých gagů, takže záleží na každé slabice, lexikální volbě i pořadí slov, aby to čtenáři aspoň trochu cvrnklo do koutku tam, kde absurdní humor třeba přechází do trapnosti.
Hrát se tu dá i s idiolekty: venkovanům, z nichž si Ibáñez rád utahuje, dávám do pusy prvky hanáčtiny. V prvním díle jsem se u záporáků Übermenschů hodně bavila germanismy, v pátém mám zase mafii, tak jsem trochu hrábla do italštiny. To mimochodem šlo mnohem víc ztuha, čeština má víc přirozených způsobů, jak v textu parodizovat němčinu než italštinu (pokud už italštinu úspěšně parodizujeme, tak většinou intonací, což se v textu dělá blbě).
Jsou to každopádně takové jazykové hlavolamy, ke kterým se s radostí uchyluju ve volných chvílích. Tak snad to ještě někdo ocení, třeba aspoň ty děti. 😊 Čtvrtý díl Kdopak by se býka bál vyšel nedávno v nakladatelství Crew.
Am I deciding literally two days before November that maybe it's time to do Novel Month again? Maybe? Because I have literally hundreds of other more important things to do, and you know I'm going to avoid doing critically important things in a timely manner by any means necessary.
Also, still hilarious that creative people hate AI so much that promoting it literally killed NaNoWriMo instantly and people just kept on doing the thing under another name.
@matt I think that the people who write novels for Novel Month were entirely in the right to shun the NaNoWriMo folks for selling out their community to AI grifters.
Whether there is a valid role for AI in creative writing, perhaps the jury is still out. But, I hate AI-generated prose. Not merely dislike it or find it worrying for ethical reasons...I hate reading it. So, if AI-generated prose is to be a part of writing novels going forward, I think those novels will probably suck.
Archos
in reply to Zloběna • • •Sensitive content