I so badly want to see this site to take off.

If you're #ActuallyAutistic, you know just how much of a problem these things can be. For me personally, having the information about lighting and how to get service in advance would be a huge help.

prepped.to/

If you see an invitation to participate in a trans or genderqueer study sponsored by Northwestern University, steer clear. It's a rigged, unethical scheme to undermine access to care. Spread the word.

#LGBT #LGBTQ #LGBTQIA #trans #transgender #TransRightsAreHumanRights

This entry was edited (1 day ago)

RE: tomkahe.com/@tom/1159910433638…

We should really get @matrix, @delta @swf @Bonfire and @xmpp people in a room to discuss how to move e2ee overlap work forward this year. It's time...

in reply to wakest ⁂ FOSDEM

fwiw some of us just launched autocrypt2 today, a new transport-independent e2e scheme to do PQC and reliable deletion (forward secrecy), using clock based instead of message based key ratcheting. ietf.org/archive/id/draft-auto…
This entry was edited (21 hours ago)

also, as much as my soul hurts a bit for all the AI spending this month, this is the exact insight I needed for my main job around AI. Now I know which one to turn to, as on the enterprise-level I've always had access to these models, and know which to use for the right kind of problem solving at work. To me that's huge, even if I'm out a lot more money this month. I don't think a good engineer is quite beholdened to a single AI tool, even if you use one of them at the power (max) level and the other at a much lower "per-usage" way when you need insight one LLM just might lack. But again, the fact that Open-Claude got like 800000 stars in a week is nuts and really shows I'm not the only person thinking this way either.

:blobcatbusiness: What's the most common complaint I've heard about Linux?

Not the installation process.
Not finding a distro.
Not getting programs to work.
Not troubleshooting.
Not hardware compatibility.

The most common complaint about Linux I've seen is this:
For a normal computer user, asking for help is just about impossible.

They ask a simple question and:
People respond "Did you Google it?"
People complain that the question wasn't asked "correctly".
People respond "RTFM"
People get mad??? at them for making an easy mistake.

We can't expect normal people to know to, or even know how to deal with any of that stuff.

Search engines these days are awful, manuals are hard to read for most people (especially stuff like ArchWiki), and normal people make mistakes we think are easily avoidable.

The solution to making Linux more popular is not ruthless promotion. The solution is to actually help the people who are trying to use it. :ablobcatattention:

#Linux

This entry was edited (1 day ago)

#Clankers

Sensitive content

in reply to Ethin Probst

@matt @Bri The only official clients are the NVDA controller client IDL/ACF files. Stuff for JAWS (FSAPI)/ZT and stuff are either generated by decompiling the TLBS or (in ZT's case) I used the microsoft-specific #import directive to have cl.exe generate the C++ code because midl refused to compile the OLEView-generated IDL (because, apparently, OLEView is a dumb decompiler)
in reply to Bri🥰

@Bri Rust compiles to native code like C++, using LLVM with all the optimizations. And Rust is designed to take advantage of all those optimizations. Sometimes the constraints of the Rust ownership and borrowing model mean that safe Rust can't be *quite* as fast as C++. Likewise for bounds checks, which are sometimes necessary in safe code. But those minor speed limitations are IMO totally worth it for safety, and Rust is basically in the same league as C++ in terms of efficiency.
in reply to Matt Campbell

@matt @Bri And other times, the fact that Rust is so safe makes programmers more confident to introduce concurrency and refactor more, making the program much faster. Not to mention all the optimizations the compiler can do because it doesn't have to worry about aliasing and stuff.

I watched a talk once where somebody rewrote rsync (that was GO and not RUst I believe, but a similar principle still holds), and their version was far faster than the one in C, despite using an older and slower version of the protocol. Not bcause the language itself was faster, but because it encouraged writing better, faster code, given finite programmer time.

I can really see why developers like Claude and that Open-Claude thing has taken off honestly. Compared to GPT's container, Claude can pull down packages, so for phoneme tuning, I can actually get Ukrainian IPA from Espeak inside it and give it guidance on what and how to tune where. Honestly, Claude's Max plan is more of a value worth than GPT's Pro plan is, and it is half the cost, so starting next month I'll save a lot by not paying for that anymore. It was just ouch for a month where I had to try the high-teared plans on each. Gemini still sucks for not letting you download patched files in browser. I don't know why Google's lagging behind on this one, yes, it can modify your code through Gemini CLI but if you drop it a file in-browser it has to rewrite the entire thing from scratch if you ask to insert a function or whatever. Nope. Also with Claude / GPT you can paste in the block you made in your IDE and say, "here's the diff, the new file, this is what we're working on."
Open AI, if you re-upload the file with the same filename, it does not override to the newest version, so you end up with something-revised-23.cpp because you have to keep incrementing the version count for each new revision you or it made. Claude can grab the latest copy from /uploads in its container and you then can stay with a clean file, but then might lose revisions that worked earlier, so I guess both can work, but the more and more I use them side by side, Claude does win.

reshared this

in reply to André Polykanine

it is. I was trying the API and I blew through $120 in a matter of about 3 days. Youch. That's wen I really considered the Max plan because I realized that the usage cap raises they give you on it meant I wasn't paying per request's tokens but per actual time usage per day/week, looks like how they calculate. So now I haven't hit it once, whether through the $100 plan (web) or through Max web or API. I'm a lot happier with it even at that "5X more usage" one. Haven't thought of "20X" more usage even which is nuts to me.
This entry was edited (21 hours ago)
in reply to Tamas G

also, I have never, ever seen Claude think for 58 to 80 minutes. That is just nuts. I do like that GPT thinks through more carefully and methodically (chunk by chunk) on a problem, Claude is a bit like the impulsive kid who can spit out a bunch of code at once, never even check its code thoroughly (sure, it'll check, "Will it compile?" but we engineers know there's more to code than that), and it can easily blow up your file. OpenAI doesn't. So I'm not saying Claude is all shining stars and shimmering butterflies. But here's how I see it.
If you're new to coding, don't know a lot of programming fundementals, the more expensive Open AI pro plan pays off. You spend less time debugging because GPT took 65 minutes to think through and carefully reflect on its work, almost like, in deep meditation.
But if you're a seasoned programmer who's familiar with concepts, logic, how to write debuggable code that's modular and you know the languages its coding, Claude will absolutely not set you back for costing half the price and being "too nimble." If you know how and what to do, it might actually make you more productive.
in reply to victor tsaran

@vick21 the web claude can compile too / build! I was surprised when it actually took my repo and ran the full build stage in its container, and spit out for me the build artifacts. Made it easier than spinning up my VM manually to build or a second machine with Linux for that, I was a bit surprised. But yeah, will have to see and test how Claude CLI is, haven't messed with it on the API level beyond some Python / Wrapper CLI tools that let you talk to Claude, but I do wonder if theirs also uses the horrible "nodeJS-packed" rewrites-terminal-every-time method as Open AI does. Hopefully it's more accessible, Gemini CLI still intimidates a me a bit too for the menu driven nature and wrapping, which is why I've stuck a lot to the web side beyond Codex on OpenAI, but then I'm restarting the session every 3-4 messages because too much gets re-written to it. Haha. I do like the idea of the AI using your installed tools since then you're working out of the same workspace, less context switching, so it's for sure the better workflow.
in reply to Tamas G

Claude CLI is just as much of a mess as all of these CLIs go. Gemini-cli has a screen reader mode which makes things tiny bit easier. It's the rendering engines that are usually at fault, like Ink, and such. I am using both Claude and gemini-cli via the ACP protocol which spits out pure JSON containing responses. The downside is that you do not get some of the cool things that come with the CLI.

A random thing I discovered yesterday on my iPhone running iOS 26: when filtering through apps in the Braille Screen Input mode on the home screen, your input is translated into Braille and retranslated into all available languages. As a result, when using the Braille table of one language e.g. Polish, you can use the dot combinations for characters of other languages and matching apps will be found. No more switching to German or Czech just to find an app with some diacritics. #Blind #Accessibility #iOS

reshared this

The very first beta version of Fedra, the completely native, blazing fast Mastodon client by @Aryan and I is here! It certainly doesn't have all the features of a Mastodon client yet, one big one we're still missing is searching, but it's stable and snappy enough that we can't resist letting people have their hands on it. Do note that because of how early of a beta this is, there is currently no documentation, but most things should be pretty self-Explanatory. Do also note that this beta will stop functioning on March 1st 2026 at Midnight UTC, but we'll likely have another beta or stable release out by then. quinbox.xyz/fedra.exe Enjoy!

reshared this

e-mail services I'm currently testing for #accessibility - proton, @Tutanota then Infomaniak and @mailbox_org

As folders and messages managing via screen reader and keyboard, MailBoxOrg, a service from Berlin, seems to be the best. It's paid only.

Proton is in the middle, while infomaniak is interesting for the amount of services it provides in basic plan and as it's affordable in price, unfortunately Tuta, which I'd have promoted as the best for first period of 2025, since summer it has became almost impossible to use through web. And phone, as it's not native, it's quite difficult to use as a blind person.

I wonder if there are even other secure, privacy-oriented, Europe-based and open source, accessibility-friendly even more of those characteristics together, for me to try.

I want GMail to stay just here for newsletters and stupid ads, but I need a very reliable service for personal e-mail messages.
Yes, I even considered self-hosting my e-mail into my web site's provider. But give to doctors, shops, tech assistance and so on, a complicated address such as something at plusbrothers dot net! Especially if they speak no english!

For that purpose now I have nickname at pm dot me, with proton, that seems the best option. But a little Trump Oriented, I fear.

@GrapheneOS Please become more active on Lemmy using your Mastodon account (All Lemmy instances provide unauthenticated full-text search, unlike Mastodon. It only lacks fuzzy finding), or set up your own Lemmy instance and account, to more effectively address mentions and false claims about GrapheneOS. You're doing a great job interacting with the community here on Mastodon!
in reply to midnightblue

Our community needs to be far more active engaging with people about GrapheneOS across platforms. Provide factual information and counter inaccurate attacks on the project and team while remaining calm and polite even in the fact of people engaging in personal insults and linking harassment content. We have a lot of users and a large community but aren't receiving enough help from them with this. We cannot be everywhere and cannot spend more time countering misinformation.

Accessibility fedi, may I nerd snipe you a little. At the makerspace we have 9 doors. I'd like to repaint each door a unique colour, so we can have simple stuff like "lock on blue door broken" and remove ambiguity. Except, as xkcd has proven, colours are hard.

Is there a set of 9 (maybe 10 to allow for future expansion) unique colours that work with all colour blindnesses, and related visual accessibility needs?

#Accessibility

in reply to Quixoticgeek

Sorry, I don't know what you are up to, but think about blind people, or, if it's not pertinent, about color-blind people. I'd suggest to proceed with painting but have other reference, like "third door" or "wooden door" or "door with a flower on it", I'm making stuff up now but color alone is a no-no per any accessibility standard. #Accessibility

@GrapheneOS Are Pixel devices running GOS vulnerable to this GNSS baseband processor tracking?
an.dywa.ng/carrier-gnss.html

My rabbit hole for today is a Youtube channel of someone who took their sweet time and not only recorded different phone announcements across the world but also went through the ITU and ETSI recommendations and actually recreated different telephony signaling sequences from different parts of the world at different times going back to 1955, then did some more tinkering and did the same with some cordless phone ringtones. Just to get a break, they simulated a couple of international calls, putting the right signals and probably TTS-created voicemail greetings at the right places. Enjoy! youtube.com/@tonsoftonz?si=yN3… #Telephony

reshared this

What the actual fuck. It's not April Fools Day, right? This is real?! Fuck you, SpaceX. Maybe this will help regulators realize how fucking shortsighted companies' plans in orbit are?

Surely the FCC won't rubber-stamp-approve this one??... ONE MILLION STARLINKS FUUUUCCCKKKKKK

au.pcmag.com/networking/115649…

Gibt es eine ernstzunehmende Alternative zu »Vimeo«?

Ich bin derzeit bestrebt, mich von einigen Firmen und derer Apps etwas freizumachen, respektive möchte ich gerne meine Daten immer mehr auf europäische, bestenfalls deutsche und schweizer Server legen.

Nun möchte ich gerne Videos meiner Bands »Bellybutton & The Knockwells" und »DramaGold« (ca. 20 GB), die bislang bei Vimeo liegen, eben von Vimeo wegziehen lassen.

Welche Möglichkeit würdet Ihr empfehlen?

#video #LeaveVimeo #music

This entry was edited (23 hours ago)

I'm fed up of standard answers such as "received, we'll pass it on" when I report accessibility issues to @Tutanota - I switched to paid subscription after they quickly enough resolved a captcha issue, but now it is at least since summer that I do not even manage to read e-mails from my web browser.
And on TUTA web page they publicly declare they respect anti-discrimination values.

I am blind, I am an accessibility advocate, no one has the right to silence me any how.

I am not saying this is something intentional or malicious, I am looking at the result. The real life result of what maybe is an unwanted effect of even a single component update.
As inaccessibility is fought through money, if within 3 months I'm not getting an effective concrete response, I'd ask for all my 36€ back till the last cent.
No time to do this yet, I want to trust them once more.

Created an issue on their github repo.

github.com/tutao/tutanota/issu…

#accessibility #a11y #blind

This entry was edited (23 hours ago)

reshared this

in reply to Elena Brescacin

Hi Elena. We're sorry this has been your experience. At Tuta we do respect anti-discrimination values, and we try to welcome everyone to use our services. Our teams are trying to improve all aspects - including accessibility issues, so we can assure you this is something we are aware of. Sometimes things do take a bit longer as we're still a growing team. We hope you can understand.

If in the EU. Try WERO and tell your bank how to make it better if you truly care about an alternative to PayPal and the Visa/Master duopoly. Don’t expect the perfect solution. Understand that alternatives are never better or perfect from the get-go but that we need to tell them through using them how to become better and better.

wero-wallet.eu

This entry was edited (23 hours ago)

:boosts_ok_gay: i wanna see which networking vendors fedi prefers
pick your favourite

  • Palo Alto (0%, 0 votes)
  • Juniper (0%, 0 votes)
  • Cisco (0%, 0 votes)
  • MikroTik (100%, 1 vote)
  • Netgate (pfSense and OPNsense live here) (0%, 0 votes)
  • Zyxel (0%, 0 votes)
  • DrayTek (0%, 0 votes)
  • Ubiquiti (0%, 0 votes)
  • iboss (0%, 0 votes)
  • HPE/Aruba (0%, 0 votes)
  • Fortinet (0%, 0 votes)
1 voter. Poll end: in 5 days

Tried to extract my own glottal pulse to make the synth sound more human. Learned my voice is too gentle for radio. Sadness fills my soul. That's probably why I didn't stick with radio shows.
I recorded sustained vowels and used IAIF (Iterative Adaptive Inverse Filtering) to extract my glottal waveform - the raw "buzz" before your throat shapes it into vowels.
What I expected: Rich, characterful human excitation to replace the mathematical model.
What I got: A softer, breathier sound than pure math! 😅
The mathematical LF model with sharpness cranked to 10 actually produces MORE harmonics than my actual voice does. That "chest resonant radio announcer" sound? That's aggressive glottal snap that not everyone has.
This entry was edited (23 hours ago)

URLCheck sits between your apps and web browser and intercepts web searches, stripping out all of the additional tracking, that is, everything that comes after the first question mark: This FOSS app is how I open links without being followed around the web makeuseof.com/foss-app-open-li…

reshared this