Euro prudce klesá po zprávě, že rezignoval nový francouzský premiér
(ČTK)
https://www.ceskenoviny.cz/zpravy/euro-prudce-klesa-po-zprave-ze-rezignoval-novy-francouzsky-premier/2730573
Takový hezký den to mohl být. Člověk ráno vstane, všude svítí — vždyť když vyhrála ta lepší vláda, elektřina bude skoro zadarmo. Možná nám i budou doplácet! Do Německa jsem to valil jen na dvojku, ať ten diesel pořádně hulí. Den jak z pohlednice… Než jsem se v práci blbě otočil — a cvak! Záda v háji. Sotva se hýbu. Tak zas marodím. Mě už fakt klepne.
důležitý je si to udělat hezký. Neboj, oni ti teď dají protekční, naprosto zdarma léčbu. A možná na tom vyděláš, když budeš chabrus na záda jak důchodce… To jsou přece ti, co se teď budou mít nejlíp.
The selection of apps being offered on @accrescent remains pretty random, with barely any infos about these provided within the app. Neither the Play Store the F-Droid approaches to this are perefect, but this is way worse. 😬 Still wondering what the target audience for Accrescent is...
Tato kniha je, kdyby to někoho zajímalo, zdarma k dispozici na webu Městské knihovny v Praze. (Ovšem bez obrázků, pokud se něco ohledně práv nezměnilo.)
🇪🇺 If you already contacted your MEPs: Contact your local media TODAY to ask them to report and inform on the dangers of Chat Control.
🌎 If you are not European: Spread the word! Talk about it on social media. Contact your local media. Contact your representatives to share your concerns about similar proposals and the loss of privacy rights. This would affect you too!
This is an urgent matter. We cannot let the West slip in this authoritarian dystopia. We must stop Chat Control ✊🛑
📣 Do-It-Blind (DIB) online Besprechung am Montag, 6. Oktober, um 19:00 Uhr. Du bist eingeladen! bbb.metalab.at/rooms/joh-szv-o… Wöchentlich am Montag besprechen wir neue Formen der digitalen und inklusiven Zusammenarbeit. Mach mit! 🛠️ #make #blind #inklusion
Learn using BigBlueButton, the trusted open-source web conferencing solution that enables seamless virtual collaboration and online learning experiences.
I'm sure I recently saw a post to the effect that when a measure becomes a target it stops being an effective measure - or something very much like that. Now I can't find it. Does this ring any bells for anyone? 🔔
I feel like WhatsApp and Facebook messenger need to have a way to disable calls entirely. Why? Because some humans are the reason I need to enable do not disturb. humans need to learn some manners.
Das Client-Side-Scanning ist ein Angriff auf Verschlüsselung und untergräbt jegliches Vertrauen in private Kommunikation. Die Bundesregierung muss sich endlich eindeutig dagegen positionieren und die #ChatkontrolleSTOPPEN!
Twitter was the place where the news happened, not where it was reported. You want to be where Taylor Swift announces she got engaged, not where news outlets report on her engagement. Bringing news outlets onto the Fediverse is not going to recreate the magic that Twitter had.
True. Yet it's a critical mass problem. Taylor Swift will join the Fediverse if enough people are here. News outlets may help to pave the way. Or even better, the journalists not the news sites themselves.
@lostgen yes, public broadcasters have a moral obligation to be here. Cross posting an RSS feed onto the Fediverse is not going to move the needle. Original, interesting content will.
I was at #FediDay25 in Berlin over the weekend. Lots of discussion about how we can get institutions and public broadcasters onto the #Fediverse.
When I was on Twitter, I mostly followed personal accounts from journalists and researchers because I was interested in their hot takes that didn't get filtered through social media departments. If we want to get them on the Fediverse, it needs to become a place where they genuinely want to be. Public pressure won't be enough.
No shit Java can't get the HiDPI settings properly.
xdpyinfo reports my screen is bigger than it is. dpi is 256, not 96 like it claims. (not sure what Java does to be honest. so maybe it's a red hearing)
just finished a fabulous book by Steven Greenblatt describing the life of Christopher Marlowe. Quite an amazing read and lots of insights into renaissance England. Highly recommend. oh yeah, it was read by one of our favorite narrators Edoardo Ballerini! 🙂 audible.com/pd/B0FG972PJV?sour…
Check out this great listen on Audible.com. The Pulitzer Prize–winning author of Will in the World reveals the daring and subversive life of Christopher Marlowe―Shakespeare’s contemporary, inspiration, and rival.
* Vaguely radio-looking. * Small, portable, like a Bluetooth speaker. * Connects to a home server with music. * Lets one choose music through a screen and good controls. Or a touch screen or something.
Basically I want the convenience of a Bluetooth speaker without having to fuck around with a phone. I want a dedicated radio-like box that knows how to talk to my home music server.
@vivia any server, really... I've been using navidrome, which works fine with a web client, but all the subsonic apps I have found are absolute garbage. I haven't tried musicpd but have had multiple recommendations. Also haven't tried jellyfin.
check out some of the products from Ocean Digital. I have a couple that are many years old but they still make me I've I think.
Had tried many before I kept these. They do Internet radio with simple presets and also will connect to home server eg NAS (via DLNA) for playing playlists and/or selecting by artist, album, title.
Probably other similar products around but have been mostly happy with these.
Kinda funny how easy it is to complain about something when you're not the one having to actually deal with the core of the problem. And just leaving that there.
We went from "we will deport the worst of the worst" to "we will deport anyone here illegally" to "we will round up anyone who looks wrong" to "people who live in cities are the enemy" essentially overnight.
Wow, what a great use of MCP for interacting with Chrome Devtools: addyosmani.com/blog/devtools-m… I tried this with gemini-cli and it works quite well. Obviously, any other MCP client will work!
I keep seeing things about the big Audacity redesign. My only question is: how accessible will it be? As I recall, there were accessibility regressions in the past that were not addressed. I know Reaper is the one everyone likes these days, but I have a soft spot for Audacity. It was the first editor I learned to use. It taught me a lot of core concepts. I don't do much audio editing these days, but I still hope Audacity becomes a solid option for screen reader users.
It will be, I can pretty much guarantee that. The company behind it also develops MuseScore, and when version four came out, they did a huge accessibility push. The UI engine they're using is the same one that will be used in Audacity, and I would be extremely surprised if accessibility was not heavily considered during this redesign. I think that's one of the problems they're trying to fix, as right now at least on macOS accessibility with the current version is pretty bad.
This video digs into the research and design that went into gradually transforming Audacity 3 - a free and open source audio editor and production app - and ...
Well I have done it! I am now taking a personal #NVDA challenge. I have just uninstalled #JFW from my computer. I am going to see if between now and November 4th I can make it without using JFW at all. I already know of one area that I am going to have troubles, so if anyone who knows how to write things for NVDA wants to try to help, this is your chance. If someone could write an addon for NVDA that will allow someone to go to a web site, and save it as a favorite into a folder on there computer like #Leasie for JFW can do that would be greatly appreciated.
We look forward to hearing how you go! I know the favourite-saving issue has been a big one for you, so do please let us know if you find a solution - and of course, do let us know of any other niggles or problems which crop up!
@NVAccess I do have a salution. someone wrote an addon for me that does exactly what I wantted, and needed. I am sending them a donation as soon as I can to thank them for writing it. I am also trying to talk them into releasing it for the NVDA app store. they said if they do that they would want to put a bit more work into it, but me personally I think it is perfect. smile.
What a busy day! Yeah, Saturday ended up being a work day. Besides purchasing the AirPods Pro 3 and selling my 2 Pro, I also moved my work machine from NVDA alphas to the stable release, version 2025.3. I usually run NVDA alphas even on my work machines since I can easily downgrade if something goes wrong, and I can usually live with a few add-on incompatibilities. But with the 2026.1 alphas going 64-bit only, a good number of add-ons have been affected, including all the Eloquence and IBMTTS ones. I just can’t handle an Eloquence-less world yet, LOL! Also the horrible on-device image captioning! Let's just call it an experiment, and leave it at that. Unless the underlying model gets altered, it will disappoint many users down the road. @NVAccess
We appreciate all the constructive feedback on the quality of the image description, and yes we got that loud and clear, thank you. Note that regardless of the quality or what we do with it - that doesn't stop you using the various add-ons which provide image description (notwisthstanding that they won't have been updated for 64-bit yet, but that's expected).
You're welcome, and I do appreciate your stance on that! Of course, I'd rather see something worthwhile coming out of your implementation of the feature as a lot of time and engineering effort have been dedicated to it. But with offline AI models especially smaller ones, it would be tough to get anything useful.
"Online models often use higher processing power and can give better descriptions than local models. We have opted for local models for freedom, ease of use and data privacy reasons. Using online models generally require you to pay to use them, and set up API keys. Local models ensure your data is never sent to a third party, usage is free, and you don't need to set up keys." 1/2
2/2 "The model we are currently shipping with is Xenova/vit-gpt2-image-captioning, however we plan to add the option to download more models. Newer & better models are always coming out, so we've designed the code to be easily upgraded... We expect the quality of local models to continue to improve to a point where they are competitive with online models. I'm not aware of a clear statistic of accuracy rate for this model, but with any models, expect accuracy to vary."
Thanks for the clarification. Honestly though, in my experience (and from others I’ve seen testing it), this model hasn’t produced a single accurate or useful caption. The outputs are extremely short, generic, and often just plain wrong. I understand the code is designed to swap in better local models as they appear, which is great — but at this point it’s hard to be optimistic about offline models. The quality gap compared to online ones still feels very large. Hopefully future updates will narrow that gap.
not that I have much to contribute to this, but I know I am running OLlama on my laptop, and I connect to it from my desktop, and it has an offline moddle. it is a pretty big moddle, I think, but it can get pretty detailed if I ask it to. I don't know anything about moddles that are under say a gig in size, as I am pretty sure the moddle I have running is over a gig. I don't know how large it is though.
@JamminJerry I do prefer online models for accuracy. Yes - with online models we may face security risks. But it's up to us to decide which images to share with online models and which images not to. And why should I dedicate over a gig of SSD and vast CPU/ GPU capabilities to that when online models can get the job done? It's a huge waste IMO. Of course, it's just my preference. @NVAccess
@JamminJerry Then an offline module may never meet your needs, and that's ok - the good news is that NVDA HAS been able to meet your needs for online image description for over six years now :)
I love voice control. I hate how it takes my battery though. So I tied it to my action button. One click toggles it on. Then I do what I need to do. And I turn it off again. Another option is to leave it in command mode. That way, it's not constantly listening for dictation, but you can still use dictation to dictate. Additionally, you can turn off commands you know you won't use. I hope this helps someone who may be struggling to keep up with fast moving ableds. Happy Sunday, or Monday, as the date line flows.
Oh wow, I never had issues keeping up with humans especially in text of the sighted variety. :) That's interesting though, never used voice control glad it helps you out.
So I bought the Apple watch ultra 3 on Thursday and they said it would behere between 17th and 21st October, then I got another Email saying it would be here between 13th and 16th, just got one from DHL delivery tomorrow.
Prime example, I'm going to bed now at 22:37 with 74% battery, having taken it off-charge at around 9 this morning. I sleep with it on every night and will probably wake up with over 50%, probably over 60.
Hell yes! 🧑💻🖥️🛟 Quote from @a11yChief: Imagine you go to a site you know, and their latest update has broken accessibility. You can’t pay for what you want, or find what you need. Now imagine someone built you a panic button you could press that lets you rant about what’s broken and why, so other people with more spoons can advocate on your behalf. It lets other screen reader users know there might be issues, and to stay away. A hotkey, a button or two and an angry smash of the enter key.
I forgot how painfully slow it is to unzip a 128MB file on a FAT32-formatted SSD card. I hope the world will never get back to othat point. No nostalgia, please!
@stalwartlabs PLEASE remove the minimum 25-seat requirement for a license. There are a ton of people who would be interested in running Stalwart and desire some of the Enterprise features, but if you don't make it easy for hobbyists and self-hosters to champion your software they won't be able to speak from experience.
You can fund open source without requiring users fork over $60/yr minimum. That's what I'm currently paying for Fastmail. If I have to pay $60, I might as well keep using Fastmail which will also give me premium support whereas your $60 license will not.
I know there's a trial license option, but that isn't sufficient. Extend the olive branch by making the first tier flexible at 1-500 seats instead of 25-500.
Restaurace, která si za svíčkovou účtuje lidových 399,- korun má jako hlavní tahák na stojanu s menu CASH ONLY. Myslím, že v tomto případě není jediná pochybnost o důvodu tohoto opatření 🤷♂️
Jan Dytrych🇨🇿🇺🇦
in reply to Archos • • •Archos
in reply to Jan Dytrych🇨🇿🇺🇦 • • •Pepik Hipik
in reply to Archos • • •Archos
in reply to Pepik Hipik • • •