I'm helping an elderly person with a patient portal, and wow, there are about 4 different problems ranging from unclear instructions to pages that don't work well when the phone is zoomed in enough for an older person to actually read the text.
Plus, the iPhone keeps popping up unhelpful gunk.
This person isn't particularly afraid of technology either, but they literally can't do it.
I now think every web developer should be forced to walk through their processes with an 85-year-old.
*Round of applause, some cheerful whistles and yelling can be heard* Welcome to the club of people needing accessibility! We struggle with such things every single day. I mean, "we" as in "blind people" even more, but still 😊 On a more serious point, try reporting the issues to the dev. As you are in the US, your strong words to mention are ADA (Americans with Disabilities Act), Section 508 and WCAG.
Accessibility is not an excuse. For some backstory: I was at lunch with my parents when they told me I could do classes online but would most likely have to pay for it to which I said, "I would have to do some checking to make sure the platform was accessible." My dad then said one of the worst things anyone could ever say, family or not. "Oh, there's always an excuse." Accessibility. is. not. an excuse. It might seem like it to someone who might not have to think about something like that, but this attitude is disgusting. Accessibility comes in many forms, and having to think about whether something is going to be accessible or not is so goddamn tiring and I don't like that thought being the first thing I say, but also, that's a reality I have to accept. Again I'll say, accessibility, no matter what form you're thinking about is no excuse. After this interaction, I'm left with -1 spoons and a desire to do nothing else for the day.
Sometimes you just have to jump in, yes not everything might be the best decision to make, but honestly getting a Raspberry Pi and not waiting for the perfect moment to start selfhosting when, I had better hardware, or more money, or better internet, or lived in my own home etc, was one of my best decisions 2025. I love beeing able to spin up seriously useful docker containers in seconds, and only today, I added Beszel and linkding to the list of useful tools amongst gitea, audiobookshelf, joplin. Not only does it safe money (to host a joplin server instead of paying for joplin cloud, for example) it's fun, it's satesfying, it takes almost no resources. Just had to say it again, this is good.
Honestly I'd do this more, if I had a way of backing up all this stuff. Like right now I kinda trust all the big companies to at least cover that if something happens LOL.
@johann Fair, ngl I haven't setup backups yet but it's a high priority for sure. In the end it's only running for me so that's the good part. But I'll probably just use Kopia to backup to a Hetzner storage box, could be any other external storage place though. It'll encrypt the stuff so I don't have to worry about any company sniffing through my data.
Yeh seams like the best solution for this case, use something like Kopia or Restic to backup to remote storage. I just need to set it up properly, just have to get over the lazyness with actually looking at how to use these things.
Following the example of María Corina Machado, who gave her Nobel Peace Prize, Arnold J. Rimmer also gave his Bronze Swimming Certificate to Donald J. Trump.
Anybody have ideas about where to host a podcast (low traffic)?
One thousand years ago, I started a thing on a podcast service that got acquired by Spotify. The podcast is served from Spotify today. I’d like to change that.
I did figure out a story for hosting the podcast feed directly from my Ghost dot org site, but it’s a bit too fiddly for me to want to take that on right now, especially with a back catalog of episodes I’d want to keep online (ideally without migrating each item manually).
Hello everyone and welcome to the official #FastSM Mastodon account. Automated release announcements as well as tips and tricks and support will be posted here. I hope you enjoy FastSM! You can find the GitHub repo here: github.com/masonasons/FastSM
Is there an easy way to accept follow requests on fast SM? Example, tried hitting CTR enter like I usually do on Tweesecake, and nothing happened. I also get that this is still pretty new, so if you can't accept follow requests right now, I get it.
Today was quite productive. Got my 1st seedbox and some torrents and resilio running. Setup linkding and migrated my Firefox bookmarks, really nice tool. Some game dev here and there. The weekend was too short but wellp, off to a new week...
@menelion Resilio is crap imo but a folder I wanted to get uses it. Transmission web interface is great. Initially I wanted to just put transmission on my Pi, but honestly my network is awfull (cable internet) so I'm just sharing the managed box and with that the price with someone. I use ultra.cc as provider.
I'm a human being. I have dreams. But the genocide changed my dreams. In the past I had big dreams. But now all my dreams are to live a normal life.
What normal life means for me?
A normal life is to sleep peacefully in the night. To have a home where my family gathers. To eat healthy food and drink clean water. To meet friends in the university. That's it! Unfortunately these basics became big dreams
Please donate to help my family overcome this tough time
I'm a human being. I have dreams. But the genocide changed my dreams. In the past I had big dreams. But now all my dreams are to live a normal life.
What normal life means for me?
A normal life is to sleep peacefully in the night. To have a home where my family gathers. To eat healthy food and drink clean water. To meet friends in the university. That's it! Unfortunately these basics became big dreams
Please donate to help my family overcome this tough time
Muslim mindset: “I’m fasting, don’t eat in front of me or I might be tempted.” Christian: practices self-control and doesn’t make a public show of fasting.
Muslim man: sees a woman who isn’t fully covered and says, “Cover yourself or I’ll be tempted.” Christian man: sees the same thing and says, “I need to guard my heart and discipline my eyes so I don’t sin.”
Christianity deals with the heart. We emphasise self-discipline and self-control. Islam, on the other hand, tries to control the environment instead, asking others to change because the individual hasn’t learned to master himself.
When the heart is truly transformed, temptation loses its power. Self-control means taking responsibility for your own desires, not placing the burden on others. A disciplined heart governs the flesh, not the other way around.
What are your pain points, folks? Stuff that you hate doing or dealing with, or problems you can't find a good solution to? Stuff that other people might be frustrated with, too.
I'm looking for a way to make myself valuable to other people, as a way to both help people and also earn an income to feed my family in the process.
One thing I can do *really well* is create reliable software to automate rote tasks, generate financial/statistical/other reports, or calculate difficult solutions. Think it can't be done without LLMs? I might surprise you!
Sadly, there is no money in solving any of my problems. If there was, someone would have solved them. See, for example, my complaints about text to speech systems. stuff.interfree.ca/2026/01/05/ai-tts-for-screenreaders.html I can go into more detail about why all the options are bad if you want. But this is the sort of problem that eats years of your life, requires advanced mathematics (digital signal processing at a minimum), and advanced linguistics, on top of being a good systems-level programmer.
@fastfinge I just so happen to be an (unemployed) machine learning researcher by trade, with advanced mathematics, linguistics, and programming skills. Maybe not systems-level programming, but I could probably find someone who does that and work with them.
Given that the first two responses I've gotten were both about accessibility, there might be more of a market for this than you think, and also, it might make a good way to demo my skills even if it isn't paid work.
The reason I say systems-level programming is mostly because for a text to speech system used by a blind power user, you need to keep an eye on performance. If the system crashes and the computer stops talking, the only choice the user has is to hard reset. It would be running and speaking the entire time the computer is in use, so memory leaks and other inefficiencies are going to add up extremely quickly.
From what I can tell, the ideal is some sort of formant-based vocal tract model. Espeak sort of does this, but only for the voiced sounds. Plosives are generated from modeling recorded speech, so sound weird and overly harsh to most users, and I suspect this is where most of the complaints about espeak come from. A neural network or other sort of machine learning model could be useful to discover the best parameters and run the model, but not for generating audio itself, I don't think. This is because most modern LLM-based neural network models can't allow changing of pitch, speed, etc, as all of that comes from the training data.
Secondly, the phonemizer needs to be reproducible. What if, say, it mispronounces "Hermione". With most modern text to speech systems, this is hard to fix; the output is not always the same for any given input. So a correction like "her my oh nee" might work in some circumstances, but not others, because how the model decides to pronounce words and where it puts the emphasis are just a black box. The state of the art, here, remains Eloquence. But it uses no machine learning at all, just hundreds of thousands of hand-coded rules and formants. But, of course, it's closed source (and as far as anyone can tell the source has actually been lost since the early 2000's), so goodness knows what all those rules are.
@fastfinge Reading your linked article article and this reply, I get the sneaking suspicion that HDC (hyperdimensional computing) or other one- or few-shot learning methods that are designed to factor the model into independent components that can be quickly recomposed in new ways might be appropriate. The idea would be to, as you suggest, learn the values for these components using machine learning, but also the mapping between them and the sounds produced, so that each becomes separately tunable on the fly.
HDC has the added advantage that it is great for working with "fuzzy", human-interpretable rule representations, is typically extremely efficient compared to neural nets, and even meshes well with neural nets and gradient descent-based optimization.
Do you happen to have data of any sort that could be used for training?
In general, for training the rules for pronouncing English, the CMU pronouncing dictionary is used: www.speech.cs.cmu.edu/cgi-bin/cmudict When it comes to open-source speech data, LJSpeech is the best we have, though far from perfect: keithito.com/LJ-Speech-Dataset/ And here's a link to GnuSpeech, the only open-source fully articulatory text to speech system I'm aware of: github.com/mym-br/gnuspeech_sa?tab=readme-ov-file I'm afraid I don't have any particular data of my own.
Sadly, this is so far outside of my expertise and abilities it's not even funny. I have an excellent handle on what's needed, and the vague shape of the path forward, but actually doing any of it is way outside of my skillset. If it was anywhere near something I could do, I would have started already.
@fastfinge my thinking is that, sure, I can build a thing, just like all those other folks, but you know the actual needs it would meet firsthand. That's tremendously valuable and can make the difference between something awesome and something completely useless
Absolutely yes to all of the above. I can think of another 10 people on Mastodon at minimum who are also ready and willing to help where ever they can. Just none of us with the skillset to do the actual work.
When it comes to requirements, in general, if it can work with both the SAPI5 and NVDA addons API, it will suit the requirements of speech dispatcher on Linux and the mac API's. The important thing is that most screen readers want to register indexes and callbacks. So, for example, if I press a key to stop the screen reader speaking, it needs to know exactly where the text to speech system stopped so that it can put the cursor in the right place. It also wants to know what the tts system is reading so it can decide when to advance the cursor, get new text from the application to send for speaking, etc. I really really really wish I had a better example of how that works in NVDA than this: github.com/fastfinge/eloquence_64/blob/master/eloquence.py
I wish it would. Unfortunately, that code is what we use to keep Eloquence alive in the 64-bit NVDA version. So it's awful, for dozens of reasons. This...is a bit clearer? Maybe? Anyway, it's the canonical example of how NVDA officially wants to interact with a text to speech system, written by the NVDA developers themselves. Any text to speech system useful for screen reader users needs to expose everything required for someone to write code like this. Not saying you could or should; there are dozens of blind folks who can do the job of integrating any text to speech system with all of the various API's on all the screen readers and platforms. But we have to have useful hooks to do it. github.com/nvaccess/nvda/blob/master/source/synthDrivers/espeak.py
@fastfinge I'm wondering if the best place to put text-to-speech processing is on the hardware of the video card. Therefore, the audio is being generated by a separate system. As long as we are blue-skying it here. therefore the system would not care what system was running, rather it could read whatever was sent to the video to display. As I'm thinking this through as I write this, obviously it would need to read descrete sections or it might be garbled gibberish, but...a thought.
@RegGuy@hosford42 The issue is that then you lose all semantic meaning. Older screen readers did, in fact, work like this, especially back in the DOS days. But also, you're conflating two different systems, here: the text to speech system, and the overall screen reading system. All the text to speech system does is take text, and turn it into speech. The screen reader is responsible for everything else. And in general, we have perfectly good screen readers. Screen readers can also, by the way, drive Braille displays and other output devices.
The sourcecode for dectalk is out there. Unfortunately, It's...legally dubious at best. It was leaked by an employee back in the day, and now the copyright status of the code is so unclear that nobody can safely use it for anything, but also nobody can demonstrate clear enough ownership to submit a DMCA and get it taken off github. GNUspeech is also pretty close to what's needed, but it won't even compile without all the NeXT development tools, I don't think. So at best all that would be is a base for something else; modernizing it would probably effectively be a complete rewrite anyway.
It looks like both DECtalk and DECtalkMini are being actively maintained, with commits as recent as 1 to 2 months ago. I was hoping the copyright for the "mini" version would be unencumbered, but no such luck. It would have to be a re-implementation from scratch using this code as a guide. That's a lot easier than implementing a new system out of nothing, though.
I also have no idea about any associated IP or patents, though. Wouldn't whoever does it need to be able to prove they never saw the original code, just its outputs? Otherwise you're still infringing, aren't you? In this regard, it's probably actually a bad thing that the dectalk sourcecode is so widely available.
And most of the commits seem to be about just getting it to compile on modern systems with modern toolchains. I dread to think how unsafe closed-source C code written in 1998 is.
If you're going to reimplement something, you might be better to go with gnuspeech, as it's known to be in the GPL. At the least, it gives you a vocal model to improve on, that was coded with open research in mind, rather than proprietary code probably written for job security.
Also, if you enjoy comparing modern AI efforts with older rule-based text to speech systems, and listening to the AI fail hard, this text is wonderful for that. As far as I'm aware not a single text to speech system, right up to the modern day, can read this one hundred percent correctly. github.com/mym-br/gnuspeech_sa/blob/master/the_chaos.txt But eloquence gets the closest, gnuspeech second, espeak third, dectalk fourth, and every AI system I've tried a distant last.
@fastfinge (context: both Aaron and I are USAians)
It doesn't help that:
1. it's 150 or so years old, so a few pronunciations have changed a bit 2. the pronunciations and spellings (and hence some of the apparent mismatches) are UK English, not US English.
At a minimum, you'll have to envision skipping "r"s after vowels at the ends of words for many of these to make sense. As for the rest, I recognized a few of those from past experience with older UK English (e.g. "clerk" with an "a" sound), but a couple left me scratching my head saying "that's how people actually said or spelled it then and there?"
@dpnash@hosford42 Right, but most text to speech systems have a UK English setting. And the mistakes they're making are on things much more basic than that. For example, far too many so-called state of the art AI TTS systems can't even pronounce "Susy", "plaid", "fuchsia", and "lieutenants".
For example, here's Eleven Labs, the billion dollar voice AI company that's supposed to replace all voice actors forever. I used the voice builder to specifically request received pronunciation. That was not at all what I got. Aside from that, notice the incorrect "tear", pronouncing "plaid" "played", having no idea that "victual" is pronounced "viddle", and a number of other mistakes. I reran it just now, to be as fair as possible. It has not improved.
Compare that with the version of GNU Speech released in 1995. It still messes up "tear" and "live". But once you get past the unnatural voice, it's far more precise. And once you get used to it, much much easier to listen to at an extremely high rate of speed (4x or more) all day. All text to speech advancement from "AI" is just the wow factor of "Wow, it sounds so human!" But pronunciation...you know, the important part of actually reading text...is either the same or worse. With five thousand times the resources.
Have your tried LibreOffice? I have read that it is accessible, but I trust real users better.
What specific features do you wish for most?
I have a feeling it's probably a big ask for a single developer, but I could at least take a look at the source for LibreOffice (unlike MS products) and see if I can add the features without retooling the whole codebase.
I just want the ability to read an manipulate stuff in financial statements where its formatted right and I can just get to the catagories I need to get to so I can put in the right numbers into the boxes where they belong so for example, balancing a balance sheet or understanding and interpreting a cashfflow statement.
@NicksWorld I would probably need to sit with you to understand the dynamics of the flow and where it gives you trouble. There are a few things to unpack here, on first reading:
* Not sure what misformatting you're finding. * By getting to the categories, do you mean navigating columns by their headers? * Do you have specific spreadsheets you are working with regularly? If so, I might be able to come up with a different way to collect and/or present the information that is more naturally suited to blind users, like a Q&A format with predetermined flow.
Spreadsheets are designed specifically with sighted users in mind, so there's an element of inaccessibility baked into them. By organizing the information into a more linear, language-based flow instead of a spreadsheet, that could potentially make the process much more natural for a screen reader, and the data could then be automatically formatted as, or loaded into, a spreadsheet. I'd be interested to get your thoughts on this.
I would be willing to talk to you about htis if you're willing to talk to me through voice, if comfortable, because I'm better at speaking than writing, especially with what I'm looking for. As for libra office, I don't reallly have experience with it, because I kind of thought libra office was libravox lol.
Sure, I bet it's a bit of a pain for you with text-based discussions! I'm awkward on phones but willing to give it a shot, if you think it's worthwhile and can put up with my spoken awkwardness and fumbling with words. (I communicate so much better when I can write! lol)
Can I suggest, though, that first it might make sense to get familiar with LibreOffice and see if it does a better job with the interface than Excel or other such software? It would be a shame to waste our time and effort on a problem that's already solved. It might also turn out that you have different pain points with the open source software that I can actually modify.
Free office suite – the evolution of OpenOffice. Compatible with Microsoft .doc, .docx, .xls, .xlsx, .ppt, .pptx. Updated regularly, community powered.
@NicksWorld So I just downloaded this out of curiosity, and it seems like they definitely did some good work on accessibility. Unfortunately on macOS when used with VoiceOver, it doesn't really behave like a standard Mac app in terms of the UI and VoiceOver doesn't work how you would expect. This completely makes sense as it's open source and likely not developed using something like Swift UI, but for Mac users I would honestly stick to iWork.
A good friend of mine needs a lot of help. Facing health challenges as well as eviction, she needs enough covered to keep her, her partner and their cats from becoming homeless. Payment is going to be due by February or they get evicted, and they need as much covered as possible. They've personally helped me out before in my time of need nearly 2 years ago; please help me boost and cover their costs, I'll be forever grateful!
- New code doesn't work, misses dependency - Dependency can't be installed with old PIP - PIP can't update itself, since it is too old - Delete PIP, download PIP installer - PIP installer is too new for old Python - download old-style new PIP installer - install new PIP - install dependency
Now I'm sitting there wondering what the new code was supposed to solve. Forgot why I ever tried to change that thing.
"Trottel" soll nicht implizieren, dass dieser skrupellose Anti-Europäer im Rang eines Fraktionsvorsitzenden, der ganz sicher Merz' rostiger Sargnagel sein wird, so einen imperialistischen Dreck nicht mit vollkommener, berechnender Absicht öffentlich raushaut.
Trottel ist ableistisch und darf von dir zukünftig gern vermieden werden. Behindertenfeindliche Wörter sind Nazisprech ❤ Und leider auch verharmlosend, denn, wie du selbst schreibst, machen Menners wie Spahn das nicht aus "geringer Intelligenz", sondern aus Berechnung und einer faschistoiden Schläue heraus. brrr.
Ha. Wanted to read my home timeline without seeing replies. Control shift F, unchecked replies, and boom! The only thing it didn't do that it was supposed to was retain my buffer position, but hell yeah. Love filtering in realtime.
Uh, do we know why $source$ is not working for post templates? It's strange hearing $source$ every time I want to see what the source of the post is, especially mainly locally.
This is what I mean: Christopher Duffley (@ChrisDuffley): @Bri Uh, do we know why $source$ is not working for post templates? It's strange hearing $source$ every time I want to see what the source of the post is, especially mainly locally. 14:51:26 from $source$.
LOL, it doesn't say something like "unknown" when it's a remote source. You'll eventually want a fallback for this. It just says $application.name$, lol.
We're having a bit of a rough weekend. Had some things yesterday that locked the system out. Nobody could switch and the fronter had no connection to our headspace. We don't assume today will be that much better, but we're hoping tomorrow and the rest of the week will be less of a mess. How's everyone else's time? As always we hope everyone is well and taking care of themselves. - Samara
@Bri Yeah, hopefully it will. Yesterday was just rough because people were being jerks. Today seems to be a liiiittle easier though, but that's also not a very high bar. - Samara
I just thought of something, since it came to light that Bluesky are supporting Trump's stupid actions and allowing ICE on there, I was kinda thinking is it even worth supporting Bluesky at this point? IDK how many people would even use it lol
Good point lol, I guess if some crazy fediverse like thing happens where someone creates an ATProto service that becomes a viable alternative to vanilla Bluesky it'll be useful then.
OK, so all I can trace from my minidumps is, my crashes are happening with how I'm calling LibEspeak.dll. hmm. This engine may not be ready for awhile. I'm going to have to break down both X86 and x64 calling conventions for Espeak. At least it's open-source so this isn't hard just more work. OK, looks like The issue is with eSpeak initialization: calling FreeLibrary in the destructor may unload the module while leaving the static variable espeak_initialized true. Huh. Reference counters, here we come.
Just realized: Whenever I read outrageous news about politics, my outrage comes second. First, my brain makes an attempt to find a perspective in which it might make sense to act like these morons do.
That‘s not healthy for my brain. But I‘ve trained myself so well that I can’t seem to unlearn the reflex.
And this is the main reason why I have to avoid news these days. Of course it’s also because of the helplessness and all the bad emotions. But mainly because „understanding“ causes damage to my brain and soul.
Ok, @x0 will also be happy to know: I added two new language settings: 1) autoTieDiphthongs When enabled, the frontend scans token sequences and if it sees: previous token is vowel/semivowel, current token is vowel/semivowel, current is NOT wordStart and NOT syllableStart (so we don’t smash hiatus) and not already tied, not lengthened, and the second vowel looks like a typical offglide candidate (high vowels like i, ɪ, u, ʊ, …) …it marks them as tied internally (prev.tiedTo=true, cur.tiedFrom=true), so timing treats the second part as a short offglide. The second setting is autoDiphthongOffglideToSemivowel. Optional, off by default. If enabled and autoTieDiphthongs is enabled, then when we auto-tie we also try to swap the offglide vowel to a semivowel: i/ɪ/ɨ -> j u/ʊ -> w - This is the “make the glide more obvious” switch. I hope these will help people.
Let's assume that I follow these #hashtags: #unix, #bsd, #netbsd etc. Then for the sake of argument I filter #freebsd. Now if a #toot contains the hashtags #netbsd and #freebsd should I see it in my timeline or not see it? Is there a #mastodon rule or is it a #client-rule?
Gosh though. People are really helping me add engine-level settings, that's exciting I guess. More settings, the better. The more we can expose through things people can tweak, great. I'll also be updating the phoneme editor later on because I like the idea of using a spin-edit box, and auto-defaulting paths, and a few things will be improved. It's also not considering rules when speaking text from language-specific data and that needs fixing. Bugs bugs.
oh good dear God. Now the SAPI wrapper isn't even compiling after I canged some pointer types, oh no. This so far is really a huge huge disaster. Claud to the rescue? ha.
Continue, or "try it now" a popup from Gmail now asks, offering to compose your next message with Gemini. I guess the tiny "x" is the "fuck no" button?
Yesterday I switched to Windows Terminal and PowerShell 7 from the old Windows Console Host and batch syntax, and I do somewhat feel like I've been asleep at the wheel for years.
Proper UTF-8 support, aliases, a profile to configure things at shell startup, command output capture, correct parsing of ANSI escape sequences... In short, things people should expect from a real shell.
Hopefully this doesn't prompt NVDA to start shitting the bed at every opportunity as it apparently does for many others.
It shat the bed for me using CMD, but did so less in terminal. I haven't fully switched though, feel like I know more old fashioned DOS commands than terminal stuff even now.
@FreakyFwoof I think Microsoft is working on it. I formatted 2 months ago and at that time I could not even use Codex interface with it, because it would get lost trying to focus on old lines. Now, it seems to work perfectly fine.
If you know me, you'll know that I'm not a friend of AI - but like the original Luddites I am not against the technology per se, but the use of it to drive an exploitative societal development.
@pluralistic has put it more eloquently than I ever could. So, read this:
Sigh. Since we added a new setting, have eurpod.com/synths/nvSpeechPlay… - especially if you speak Portuguese, it might help. Or maybe it'll screw things up so bad your language won't sound the same. Who knows. Unlucky 13. Guess I wasn't supersticious enough to skip it. Ah well. it ever sounds like there’s “no diphthong,” it’s usually because the boundary gap or timing makes the two parts separate, or the glide is too quiet. We just added a setting to skip boundary gaps for vowel-to-vowel transitions, which is basically the diphthong smoother. Dedicated diphthong phonemes are optional and mostly for extra fine control. To use this for your language, toggle segmentBoundarySkipVowelToVowel: true (also default) or false. This should give folks even more control over gaps, and you can mess around with the others in default.yaml for a given language to see if they help change prosody.
Well, no NV Speech Sapi, for anyone. it just crashes when trying to speak, silently, no logs, no dumps, no nothin'. Haha. SAPI engines are hardest to debug. So no SAPI engine for anyone.
Paperback version 0.7.0 is out, with a huge changelog! * Added table support for HTML and XHTML-based documents! Navigate between tables using T and Shift+T, and press Enter to view one in a webview. * Added a basic web rendering feature! Press Ctrl+Shift+V to open the current section of your document in a web-based renderer, useful for content like complex formatting or code samples. * Added a Russian translation, thanks Ruslan Gulmagomedov! * Added a Clear All button to the All Documents dialog. * The update checker now displays release notes when a new version is available. * Updated Serbian translation. * Updated Bosnian translation. * Fixed restoring the window from the system tray. * Fixed Yes/No button translations in confirmation dialogs. * Fixed loading configs when running as administrator. * Fixed comment handling in XML and HTML documents. * Fixed TOC parsing in Epub 2 books. * Fixed navigating to the next item with the same letter in the table of contents. * Fixed the find dialog not hiding properly when using the next/previous buttons. * Fixed epub TOCs occasionally throwing you to the wrong item. * Fixed various whitespace handling issues in XML, HTML, and pre tags. * Fixed off-by-one error in link navigation. * Fixed some books having trailing whitespace on their lines. * Fixed various parser issues. * Bookmark-related menu items are now properly disabled when no document is open. * The elements list is now properly disabled when no document is open. * Improved list handling in various document formats. * Improved the translation workflow for contributors. * Many internal refactors, moving the majority of the application’s business logic from C++ to Rust for improved performance and maintainability. Download: paperback.dev/downloads/ Sponsor on GitHub: github.com/sponsors/trypsynth Donate to development through PayPal: paypal.me/tygillespie05 Enjoy!
Fully blind software developer who loves making their own tools to solve problems. Primarily known for Paperback, an accessible and lightning fast ebook/document reader.
I work as an audiobook quality controller. My employer uses ClickUp to manage tasks. Unfortunately, the web interface is unintuitive and inaccessible. It contains unnamed elements, menus that expand in all kinds of ways, and similar issues. I wrote to their developers, but even after years nothing has been fixed. Fortunately, ClickUp has an open API. So I used vibecoding and now I have my own minimalist HTML application that displays my tasks, start and end dates, comments, and attachments, and allows me to post comments. I still can’t change task statuses yet—we’ll see if I manage to solve that with the help of GPT. Of course, this is not how a blind person should function in an ideal world, but it is still a way we can help ourselves. That said, I still need a server where my PHP scripts run; it could probably be done with Python as well, but PHP seemed simpler to me since it’s already running on my server. BTW it would be ideal if Clickup api documentation is one simple document which I can throw to gpt but it seems that chatgpt already know how to use it.
the whole ai-bro shtick about "ai democritizes art/programming/writing/etc" seemed always so bs to me, but i couldn't put it into words, but i think i now know how.
ai didn't democritize any of these things. People did. The internet did. if all these things weren't democritized and freely available on the internet before, there wouldn't have been any training data available in the first place.
the one single amazing thing that today's day and age brought us is, that you can learn anything at any time for free at your own pace.
like, you can just sit down, and learn sketching, drawing, programming, writing, basics in electronics, pcb design, singing, instruments, whatever your heart desires and apply and practice these skills. fuck, most devs on fedi are self taught.
the most human thing there is is learning and creativity. the least human thing there is is trying to automate that away.
(not to mention said tech failing at it miserably)
It democratizes it by making it available for the people who can't / don't want to / don't have the time for learning it.
We're already seeing non-programmers successfully create quite substantial coding projects with AI, to an extend which surprises even me, who was a huge proponent for AI in coding from the start.
Same applies to art, there are many people who need or want art (small business owners, hobbyist game creators, wedding organizers, school teachers), but don't have the budget for the real thing.
Of course, many artists and programmers don't want this to happen and try to invent reasons why this is a bad idea, just as phone operators didn't want the phone company to "force" customers to make their own calls, and just as elevator drivers tried to come up with reasons why driverless elevators were unsafe.
@miki without trying to convince you of anything (your stance on ai is yours, i'm not trying to change it), I can assure you that the reasons why many developers see generating production code with AI as a bad idea are not made up.
I am all for exchanging ideas between folks with different opinions, but this had to be said.
I see putting a prompt into AI and hoping that the generated code is correct as a bad idea, especially in complex apps that have long-term maintainability considerations, or when security / money / lives are at stake.
For throwaway projects (think "secret santa style gift exchange for a local community with a few extra constraints, organized by somebody with 0 CS experience", vibe coding is probably fine.
For professional developers, LLMs can still be pretty useful. Even if you have to review the code manually, push back on stupidity, and give it direction on how to do things, not just what to do (which is honestly what I do for production codebases), it's still a force multiplier.
@miki that's a reasonable middle ground we can somewhat agree on.
I haven't seen AI-generated code being the "force multiplier" some folks swear by, especially with newer things like the config changes in pipewire last year, but i guess ymmv
I think we're painfully re-learning the lessons we learned in programming over the last 70 or so years with AI, just like crypto had to painfully re-learn the lessons that trad fi got to learn in the last five hundred years.
Yes, you can 20x your productivity with AI if you stop worrying at all about architecture and coding practices, just like you can 5x your productivity without AI if you do the same thing. Up to a point. Eventually, tech dept will rear its ugly head, and the initial gains in productivity will be lost due to the bad architectural decisions. Sometimes that
sometimes that's the right call, particularly when you want to find out what users like and you don't yet know what you're building. It very often isn't.
@miki yeah once you stop caring about wether or not your code runs optimal, or reliable, or does what it's actually supposed to, or can be maintained properly, you can surely be 20x productive, you'll just regret it afterwards at some point (or someone else who has to fix and maintain this mess will)
@miki in the short term, it might look like you're right, but in the long term, letting unreliable tech do your job is a bad idea. I'm sure I don't have to give more reasons, right?
It democratizes it by making it available for the people who can't / don't want to / don't have the time for learning it.
No, I'm sorry, but it doesn't.
What it "democratises" is being an art director who commissions a machine to generate things derived from the (uncredited, un-compensated) work of others (whose lack of consent was gleefully violated).
Gutenberg democratised learning, with his movable-type press. Encylopaedias took that a step further, and Wikipedia amped it up again. Blogs and Youtube democratised the sharing of knowledge and skills. All these things have enabled people to learn how to do a thing.
But if you typed in a description and got a picture in return, you did not create that picture. You commissioned it.
@KatS It democratizes in the public transit way (by making transport available to non-drivers), not in the car way (by making it easy).
And btw: all art is uncredited and a lot of it is unconsensual. Outside of academia, it's extremely rare to credit every single influence that an artist used, down to Da Vinci or the Gregorian chants, as long as snippets significant snippets aren't extracted directly from that work, something that AI only does when prompted.
@miki @KatS we're not talking about influences, but more akin to "retracing".
Besides, there are real implications regarding free software licenses and AI generated slop, so it's not exclusively a moral dilemma, but a legal one too.
legal != the right thing to do necessarily, but mangling a bunch of intellectual property that's not yours through a statistical computer program isn't exactly comparable with an aspiring artist learning to draw.
@KatS Because I use it every day, and I can see how much it helps. And to be fair, it primarily helps people who get X done, not the doers of X. Just as automated telephones primarily help those who want to make phone calls (by making them cheaper, faster and much more convenient), not the phone operators who helped to make them in the past.
@miki Wow. It'll make for more efficient communication in future if you make it explicitly clear that you're democratising the commissioning of things, and working hard to devalue artistry in all its forms.
Talking about "democratising art" is typically read as making it easier for people to make art. This is what leads to this kind of convoluted exchange.
@miki @KatS don't get the wrong idea, this is pure curiosity.
I had the suspicion that the less someone knows about LLMs or ML, the more they think the tech is capable of, but that suspicion must be false, since you're around
@KatS The more you know about LLMs, the more "calibrated" you are about where they work (and don't work) right now. People who don't know much about them are either hypesters (mmaking a company of a thousand LLMs and firing all their employees), or LLM deniers. Both are just as crazy.
I also see not just where LLMs are right now, but where they are going. We went from coding agents being basically a joke a year ago, to them semi-autonomously solving (some) complex mathematical problems and being used for boring gruntwork by world-class, fields-medal-winning mathematicians. They can now also solve an extremely complex GPU performance engineering task that Anthropic used as an interview question for the most brilliant engineers in that discipline, *better than any human given the same amount of time*.
They're still much better at small, well-scoped and bounded tasks than at large open-ended problems, but "small and well-scoped" went from "write me a linked list implementation unconnected to anything in my code" to "write me a small feature and follow the style of my codebase." In a year. What will happen in another year? 5 years? 10 years? God only knows, and he certainly isn't telling.
@KatS look @miki don't get me wrong but any time i've tried using LLMs for my work, which isn't just some fun side project but actual production-running code, LLMs have been way too unreliable. It also resulted in me knowing jack shit about my own code, which is poison for long term maintainability.
Since these models are just statistically determining the next most likely token based on training data and fine tuning, without any actual understanding or thought behind it, I seriously can't see this tech being reliable enough one day. (reliable compared to humans, i don't seek 100% reliable in this case, natural language is too imprecise for that anyways. i would expect "good enough" as "as good as a professional in the given field")
The other part of the equation is the amount of compute and electrical energy necessary to train and operate such a level, and on that level, there's no way in hell that shit is ever gonna be worth it, financially and environmentally.
i'm not expecting the "make job for phone operators easier", i expect the "when i dial a number, it should be at least as reliable and efficient at routing it correctly as a phone operator would be".
you can call me whatever you want, even llm denier if you need to, but autocorrect on steroids isn't worth exploiting other people's work or boiling our oceans.
@KatS Autocorrect on steroids is basically GPT-3 tech. There's a lot more that goes into modern LLMs. A lot of the improvements are due to reinforcement learning, where LLMs learn to predict tokens that actually achieve some outcome, E.G. code that passes tests, answer that is judged "good" by a domain expert. There's still token prediction involved of course, but it somehow turns out that token prediction can get better scores than any human at (unseen) math olympiad questions. And people still say it's not in any way intelligent...
@miki @KatS if i memorize every possible answer to a specific test, i can pass too. doesn't mean i know shit about fuck.
There's no actual thinking or reasoning involved (and no, reasoning models don't actually "reason"), so yeah, an LLM isn't actually intelligent, it just shows how flawed our tests for intelligence are.
To get some actual intelligence, thinking or reasoning involved, I'd reckon we'd have to fundamentally change something in the architecture of LLMs, and use a fuckton more computing resources for a single model, and considering how much energy the current tech already wastes, and the whole shtick that made LLMs (and more broadly generative AI) work in the first place is "we discovered that there comes a point where the output gets better when we throw rediculous amounts of compute resources on the problem", and it's already getting super difficult to run and maintain.
Honestly, either you're unreasonably optimistic, or you've never taken a look at how things actually work under the hood, but I really recommend you to take a closer look at the technology you praise so much.
A couple things you could take a look at (without an AI summarizer, otherwise you'd learn jack shit):
Honestly, before you make any claims about where the tech could be and what it could do, you should have a look at how things actually work under the hood and have a rough idea of how things work, otherwise, no offense, you're just talking out of your arse.
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism.
@KatS I have very specifically said "unseen questions."
If memorizing answers was a viable strategy to pass that test, humans would have done so.
If you still believe that there's no possible use for a tool that can get gold on a never-before-used set of math olympiad question given a few hours of access to a reasonably powerful computer, and that the existence of that tool will have no interesting impact on the world... I don't know what to tell you.
@miki @KatS > If you still believe that there's no possible use for a tool that can get gold on a never-before-used set of math olympiad question given a few hours of access to a reasonably powerful computer, and that the existence of that tool will have no interesting impact on the world...
How reliable is that source? And if that's true, is it really reasonable to bet everything on this, and let this do all your work when a) you end up completely dependent on the tech and b) utterly destroy the environment in that process?
Real world problems may be less complex but might require much more context.
Oh, and don't get me started on accountability. There's a reason why curl is closing their bug bounty program.
@KatS Curl is closing their bug bounty program because it's far too easy to use LLMs to produce slop. It doesn't mean you can't use LLMs to produce non-slop, just that it is a technique some people have found to get money with not too much effort, and we haven't yet sufficiently adapted to it. This is a genuine problem.
@KatS Nothing is ever gonna work right, not even humans. Different technologies are at different points on the price-to-mistakes curve, our job is to find a combination that minimizes price while also minimizing mistakes and harm caused.
E.G. it is definitely true that humans are much, much better psychologists than LLMs, but LLLMs are free, much more widely available in abusive environments, speak your language, even if you are in a foreign country, and work at 4AM on a Saturday when you get dumped by your partner. Human psychologists do not. Very often, the choice isn't between an LLM and a human, the real choice is between an LLM and nothing (and the richer you are, the less true this is, hence the "class divide" in opinions about tech). And I'm genuinely unsure which option wins here, but considering the rate of change over the last 3 years, I woulndn't bet towards "nothing" winning for long.
Bubu
in reply to Dickenhobelix • • •"Dit is meene erste Schreibmaschine - ick übe noch"
Ein paar Fehler haben sich hier wohl noch eingeschlichen
.