I told GPT Pro mode that zombies ate the GitHub servers and it needs to rebuild me an entirely new set of phonemes in phonemes.yaml because I've lost all copies and only have the language packs. I told it that if this doesn't get done in the next hour, the zombies will come for the language pack files and eat them too. Let's see what comes of it while we reply to Slack messages.
This entry was edited (4 hours ago)

So I left for the weekend to visit family, came back -- avoided contact with sick people the whole time -- only to find my wife was sick with a cold and she didn't tell me and now I have it

Just pisses me off because I've been able to go a couple years without getting anything and now this. It always ruins my week. I can't get shit done because it's a constant nuisance disrupting my concentration

People who are coming to the #FOSDEM this week-end : please do not forget to wear a facial mask to protect yourself and others. Flu this year is pretty agressive and the orange code for respiratory infections has been raised by the Risk Management Group.

#retoot can save lives.

#brussels #respiratoryInfections #covid #flu #mask

This entry was edited (7 hours ago)

RIP Gladys Mae West, the Pioneering Black Mathematician Who Helped Lay the Foundation for GPS

openculture.com/2026/01/rip-gl…

FYI, I reviewed my account on dragonscave again: @jonathan859@dragonscave.space, this is not an advertisement, but as I'm feeling rather emotional again I'm shifting the kind of personal ranting posts and also other stuff that's not my typical tech timeline to there. Also to filter who sees what, in terms of post privacy etc. I'll probably share more occasional stuff like music I like or other stuff I don't wanna spam to uninterested people as well. Looking forward to nice interactions.
This won't change a whole lot for this account over all.

Habe beim "Entrümpeln" noch einige DVD´s und Blu-ray´s gefunden, die ich gegen Übernahme der Portokosten abgeben möchte. Wer mag darf gerne eine kleine Spende dazulegen, aber das bleibt euch überlassen.

Die DVD´s und Blu-ray´s werden als ein Paket abgegeben. Der Aufwand ist mir zu hoch sie pro Stück zu versenden.

Ein Boost wäre nett, Danke !

#verschenken #dvd #bluray #boost

This entry was edited (11 hours ago)

In 2026, it'd be nice if we could scrap euphemistic terms like 'special needs' 'Additional needs' 'different abilities' 'very special kids' or anything 'inspo-p0rnish such as 'Autism is a superpower!'. Instead, use the words 'Disabled' 'Disability' ... it's not an offensive term. In fact, many disabled people find these sugarcoated terms actually offensive. And let Autistic and other neurodivergent people decide if they have exceptional powers, or are disabled by societal structures. Or both at once.
Thanks :) 🙏💪 ❤ #disability #neurodivergence #ableism

So what is development like for the BTSpeak?

Are you limited to making apps that only work with desktop mode, or can you do things that also integrate with the non-desktop experience? (Not up on my BTSpeak terminology so maybe there's a better description for "non-desktop-mode experience" but hopefully what I mean is clear)

What is development like? I know there are Python tutorials but is it limited to Python? Or can you build a TUI-style app and adapt it to work with specific keycodes/conventions?

Are there any developer docs?

I've had an idea for something that would likely be great for the BTSpeak and other notetakers for a while now, and my assumption is that the BTSpeak is probably the most open/easy to develop for. If that's correct, I'm curious what that process looks like.

On this day, forty years ago: 28 January 1986.

I was working on my astronomy PhD in the terminal room at the Royal Observatory Edinburgh.

Someone came in & told us the awful news. After so many launches & astronauts, we’d grown blasé & didn’t pay much attention anymore.

That changed in 73 seconds on that cold day & we learned again that space is hard.

I still remember their names:

Onizuka, Smith, McAuliffe, Scobee, Jarvis, Resnick, & McNair.

Ad astra, STS-51L Challenger crew ✨

#space

This entry was edited (11 hours ago)

Most hashtags are pretty obvious, you just put # in front of a topic. For example if a post is about coffee it might have the tag #Coffee

However, many popular hashtags have difficult-to-guess names, so I've started a list to help people discover them:

➡️ fedi.tips/fun-and-useful-hasht…

Copy-paste a hashtag into the search box in Mastodon if you want to browse or follow it.

If you have suggestions for non-obvious hashtags that should be added to the list, let me know in the replies 🙂

#FediTips

If you would like to help financially support the development of a new Wayland compositor for Xfce, you can do so by contributing funds via our #OpenCollective accounts.

For US contributions:
opencollective.com/xfce

For EU contributions:
opencollective.com/xfce-eu

These contributions will help pay for the funding of longtime Xfce core developer Brian Tarricone to create xfwl4, a brand-new Wayland compositor for Xfce.

Thank you for your support!

#Xfce #Wayland #Rust #FOSS #Linux

From the Slint email list:

Dear all.

I am very sad to inform everyone that our friend Didier died last week.

Early 2015, I asked on the slackware list if brltty could be added in the
installer ; Didier answered promptly that he could do it on
slint. Afterwards, he worked hard so that slint became as accessible as
possible for visually impaired people.

You all know that all these years, he tried and succeeded to answer as
quickly as possible to our issues and questions.

He will be irreplaceable.

@Friendica Support

There is a fork of Mastodon available, called "Glitch Edition", that has some special features implemented. My new instance at social.defocn42.net is running this fork.

I came upon an interesting error when trying to connect to the instance. Adding the user "m1rk0@social.defcon42.net" failed because of an unknown network. Then i tried it the other way around from the Mastodon instance and could follow myself on Friendica. The connection is now mutual and the Mastodon instance is recognized as "glitchsoc".

Are there known issues with glitchsoc and Friendica? It's weird, that the connection can be established only from one side and is working normally afterwards.

Looks like the UK does not have our back:

"Keir Starmer rejected his Canadian counterpart’s call for mid-sized countries to band together in the face of unpredictable global powers — and insisted his “common sense” British approach will do just fine."

Like I said, we'll find out who our friends are.
politico.eu/article/keir-starm…

#CSS `@custom-media` available behind a flag in Firefox Nightly

Lovely syntax, just lovely.
nerdy.dev/custom-media

#css

Hahahaha. What if I asked an AI to rewrite the SpeechPlayer repo to work on Linux. Good dear God. I bet it would be a disaster full of bugs. I'm tempted though, just make it a new branch first and go from there. Also building a speech-dispatcher module for you Linux nerds. Really, really don't know how that will go. Now that's vibe coding to the max, I have zero familiarity with Linux argtypes and such when avoiding Windows headers. Like, zilch.
in reply to Matt Campbell

@matt oooh yeah, but that's even a bigger refactor, moving away from C++ to it, but it's also a less mess for cross-compiling. I also have no idea if there's a different UI toolkit I'd need to use for things like the phoneme tool, as it's written in basic win32 for now, something I had familiarity with from college and high school days, but also like, locks me into C++ for it pretty bad, so a cross-platform UI solution would be needed before it got too many more dialogs and settings. I could have used Python / WX, but then of course you're including all the precompiled Python runtime and can't make it as lean.
in reply to Matt Campbell

@matt I think I could do the Rust migration in layers. First, get Linux cross-compiled for the frontend and player, maybe editor, + speech dispatcher module that's simple. Then it can be re-written so the SpeechPlayer and Frontend go Rust first, then the other parts. UI development feels like the worst place to start learning how Rust is. Rust has a very different mental model (ownership, borrowing, lifetimes).

NV Speech Sapi does not have an official distributed release and probably will not until it breaks out of ESpeak's phonemizer. Sorry y'all. But because we link against libespeak, it's technically a big no-no to even have it around. Perhaps that could be an experimental project, to rebuild NVSpeech Sapi with Gruut, but damn, a 700 megabyte SAPI engine? people would cry.
in reply to Cleverson

@clv1 hmm. this is where the new settings might help you.
Mess with these settings and see if they can help in your YAML:
settings:
# Smooth big formant jumps at boundaries (good for word-initial L+vowel clicks)
trajectoryLimit:
enabled: true
applyTo: [cf2, cf3]
maxHzPerMs:
cf2: 16
cf3: 20
windowMs: 30
applyAcrossWordBoundary: true

# Give /l/ a soft onglide (reduces the “snap” at the start of a word)
liquidDynamics:
enabled: true
lateralOnglide:
f1Delta: -50
f2Delta: 200
durationPct: 0.35

in reply to Cleverson

@clv1 hmm. Try to lower: semivowelOffglideScale by a bit to see if that improves it. If it’s inside one word (like beira, loiro, ouro), start with semivowelOffglideScale. A tiny "syllable break" can happen when a semivowel is followed by a vowel, or a liquid/tap/trill, and this setting shortens the semivowel only in that pattern.
If it happens across NVDA “chunks” (label/role/value stitching), then it’s a seam, not a glide problem. In that case flip segmentBoundarySkipVowelToLiquid so the engine doesn’t insert the segment-boundary gap when a chunk ends with a vowel/semivowel and the next begins with a liquid/tap/trill.
And even when timing is right, /r/ can feel like a hard wall unless it has motion, so enabling liquid dynamics (rhotic F3 dip) often makes the landing smoother.
in reply to Andrew Nesbitt

for forges, you might wish to add @Codeberg (Germany, EU). Not sure where Sourcehut sits (is it NL, @sir ?)

So there ARE alternatives. And as already pointed out in another comment by @jens , Forgejo/Gitea can be self-hosted as well. And at least for Forgejo, Federation is upcoming IIRC, to take another hurdle (separate registrations) from self-hosted installs.

But yeah, that list reads horrible, re "sovereignty" 😢

Now it's really work time. Off to meetings while I tune phonemes in the background with my tools and AI. Tools are what make the world of difference. If you don't have them, you don't have your foundation to stand on. Tools that can give you data, look at the patterns in different ways, see if they can be stress tested in other contexts, ETC. What would we ever do without darn tools.

Another phoneme editor fix: If phoneme's fields, like _isSemivowel are not defined for the given phoneme, allow you to add it by choosing it in the list and setting a value for it. This will make adding new state-variables to phonemes easier. Remember: Flags like _isSemiVowel, _isVoiced, _isStop, are for the engine to know how to treat the sound, not for the sound itself to change. These don't influence Klatt params as much as sound timing, and which rules get applied to that phoneme token when it's used.

It was interesting to read up on the AI assisted code review at lesswrong.com/posts/7aJwgbMEiK…

For context: I'm personally responsible for at least 29 curl CVEs. Out of the recent 6 CVEs mentioned in the blog post I found two. This gives me some perspective, I think.

I do not utilise AI tools in my vulnerability research. I am also fiercely critical of harmful proliferation of AI. This is due to the unsustainable way it is currently pushed, and use of as marketing ploy and gimmick rather than producing measurable benefit to users. This leads to negative impacts on economy, education & learning, not to mention impacts to nature due to wasteful use of energy.

This doesn't mean I am against AI. I have written by own AI tooling (fully local RAG with support for arbitrary number of models running on local nodes, implemented in python). I found the usefulness of such tool to be limited at best. It is somewhat useful in mass analysis of large document bases, but the level of analysis is superficial at best. These AI models are after all just language models, and do not have any true understanding or intelligence.

And here is the gist of it: The current tools are not intelligent. Understanding this limitation is the key of successful deployment and utilisation of AI tools. The tools can be useful in certain tasks, but they do not replace true intelligence.

The AI tooling AISLE are developing certainly is one of the better uses of AI, and definitely surpasses all my personal dabbling around it. It is clear that the tool does find vulnerabilities. The key question is how much hallucinations and false positives it produces: If the tool generates thousands of FPs and the true findings are hidden among them this limits the value and usefulness of the tool (of course it doesn't entirely negate it, many tools produce false positives). In short: The quality of the findings is key, and poor signal-to-noise ratio is highly undesirable.

Either way, I think there is a future for AI tools and they definitely will be helpful in vulnerability research.

I personally will keep exercising my wetware for this work, however.

#cybersecurity #infosec #vulnerabilityresearch #thoughtoftheday

This entry was edited (7 hours ago)