mastodon - Link to source
This blogpost makes an astoundingly good case about LLMs I hadn't considered before. The collapse of public forums (like Stack Overflow) for programming answers coincides directly with the rise of programmers asking for answers from chatbots *directly*. Those debugging sessions become part of a training set that now *only private LLM corporations have access to*. This is something that "open models" seemingly can't easily fight. michiel.buddingh.eu/enclosure-…

The Enclosure feedback loop

speculation about the way (paid) software development will become LLM-only
michiel.buddingh.eu
in reply to Christine Lemmer-Webber

mastodon - Link to source

miki

This goes much, much wider than programming and LLMs.

In general, the open source world looks with disdain at all kinds of automated feedback collection mechanisms, which the Silicon Valley Venture Capital tech ecosystem has wholeheartedly embraced. OSS is still stuck in the 1990s mindset of "if there's a problem, somebody will report this to us", and That... just isn't true.

What we're stuck with is OSS solutions with inferrior user experiences which nobody wants to use, instead of a compromise where OSS software collects more data than some people would have liked, but that software actually has some users and makes a difference in the world.

To be fair, there are some good arguments against this (it's much easier to protect user privacy if the only contributors to your code are employees with background checks), but that doesn't make this less of a problem.

mastodon - Link to source
Yeah, not a bad idea to release the add-on and editor together on GitHub. Link updated. If people want they can use the packs folder from the add-on included to tweak phonemes now too, so that really solves that problem. Appreciate that suggestion. Technically it's still not bundling libespeak or linking directly to it so I'm able to do that fairly.
github.com/tgeczy/NVSpeechPlay…
Remember, the driver file is like 70 KB because it has so much code for settings, compatibility with older NVDA back to 2023.2, voice keys, all of it adds lines. I think we're good on frontend passes for now, too many settings and people can now build proper languages with this set of them.

Release NVSpeech Player with phoneme editor and NVDA Addon version 1.7 · tgeczy/NVSpeechPlayer

Many changes, and first one with the NVDA add-on bundled as this uses built-in Espeak, not code-linked: Adds proper focus management for the editor, so alt tabbing away and back, or opening the wi...
GitHub
This entry was edited (4 minutes ago)

mastodon - Link to source
I didn't put in the new phonemes into the editor yet. Going to give it a bit of time to be in the driver first, then yeah if people aren't overly upset, we'll move them into Master. For now it's a bit experimental. Maybe this afternoon. I'll move them into the master branch now, so people can see and mess with them, compare the diffs if they want, and then we'll roll it into the editor version later.

mastodon - Link to source
New NVSpeech Player build for NVDA:
Mostly language tuning. Words like "square", "fair", "bear" get proper US and UK pronunciations. Protein, Fourteen get a slight tweak, though not perfect. Some of this is just on Espeak being used for IPA and phonemizing. We also fixed it so that the "google / able" rule doesn't bleed into words like real. Overall, no phoneme changes from last night. People were not super overly mad at them getting a slight softening. At 16000, it sounds really good to me, and less sharp. At 22050, that's your own damn problem. Formant synthesizers lose their bandwidths at higher sample rates because they simply cannot fill it all in and compress their voice in an equally filling way across all frequencies. It's a thing with all of them. Eloquence at 22K? I'm realizing, not the greatest. So if you're using it at that and think, "oh now it's too soft", at least it also doesn't sound like it does at 44100 when it's at 22050, so you should really be thankful instead. Just saying. Besides, we only really changed like 17 phonemes to get this effect, it's not like all of them got a rewrite.
eurpod.com/synths/nvSpeechPlay…
in reply to Charlie Stross

mastodon - Link to source

JdeBP

Not the specific claim, but there is news coverage claiming that opposition to U.S.A. Immigration and Customs Enforcement is being suppressed on #TikTok.

Standard caution about question headlines applies, of course.

forbes.com/sites/conormurray/2…

cosmopolitan.com/politics/a701…

But let's put it this way: the notion has reached Forbes and Cosmo, and isn't just some anonymous poster on Twitter stirring the pot.

#USPolitics #fascism


Is TikTok Censoring Anti-ICE Content?

Many users are raising concerns about the platform's content moderation with its new ownership.
Annabel Iwegbue (Cosmopolitan)
#tiktok #fascism #uspolitics
This entry was edited (15 hours ago)

mastodon - Link to source

2021: OpenAI is the most expensive loss-making startup in human history, and their only product is a chatbot.

2022: OpenAI is the most expensive loss-making startup in human history, and their only product is a chatbot.

2023: OpenAI is the most expensive loss-making startup in human history, and their only product is a chatbot.

2024: OpenAI is the most expensive loss-making startup in human history, and their only product is a chatbot.

2025: OpenAI is the most expensive loss-making startup in human history, and their only product is a chatbot.

2026: OpenAI is the most expensive loss-making startup in human history, and their only product is a chatbot.

Much innovation. Very business.

mastodon - Link to source

Data control = identity control. 😨

Nubus by Nextcloud partner Univention is an identity management system that puts IT departments back in charge of storage locations, access rights, security policies, and integrations.

Discover more👇 biometricupdate.com/202601/uni…


Univention’s Nubus promises data sovereignty at fraught moment: Nextcloud

Nextcloud questions the wisdom of relying on major cloud platforms like Microsoft Entra ID to manage user accounts, access and authentication.
Joel R. McConvey (BiometricUpdate.com)

mastodon - Link to source
NVSpeech Player with phoneme editor version 1.7
• Adds proper focus management for the editor, so alt tabbing away and back, or opening the window, will land you on the first field in the window.
• adds keyboard shortcuts for commands available in the main area, so you're not left tabbing through.
• Adds new settings available in the frontend for the passes.
• YAML now gets ordered correctly as the NVDA driver writes it, so the phoneme editor will create a structurally sound form that's consistent.
Still using ESpeak as a non-linked, separate phonemizer, or you can specify your own if it runs through STDIN / CLI, as always. Enjoy!
github.com/tgeczy/NVSpeechPlay…

Release NVSpeech Player with phoneme editor and NVDA Addon version 1.7 · tgeczy/NVSpeechPlayer

Many changes, and first one with the NVDA add-on bundled as this uses built-in Espeak, not code-linked: Adds proper focus management for the editor, so alt tabbing away and back, or opening the wi...
GitHub
This entry was edited (13 minutes ago)
in reply to Alex Chapman

mastodon - Link to source

Tamas G

@alexchapman there it is, I updated the release. github.com/tgeczy/NVSpeechPlay…

Release NVSpeech Player with phoneme editor and NVDA Addon version 1.7 · tgeczy/NVSpeechPlayer

Many changes, and first one with the NVDA add-on bundled as this uses built-in Espeak, not code-linked: Adds proper focus management for the editor, so alt tabbing away and back, or opening the wi...
GitHub
@Alex Chapman

mastodon - Link to source
So HackerOne published a blog about how organizations should deal with AI slop reports. TL;DR: Use AI to evaluate them. Now we've gone full circle.
I bet @bagder has opinion on this. Also the timing seems weird, with cURL very publicly scrapping their bug bounty program.
hackerone.com/blog/ai-slop-exp…

Best Practices to Beat AI Slop With Exposure Management | HackerOne

Vulnerability volume is rising fast with AI, and triage is breaking. Exposure management and CTEM keep programs focused on fixable risk and high-signal findings.
HackerOne
@daniel:// stenberg://
This entry was edited (45 minutes ago)

mastodon - Link to source

RE: mastodon.au/@TheVoiceGuy/11596…

I have figured this out.
I had to save the HTML code in plain text instead of rich text and add the HTML extension.

Michael Marshall

2026-01-27 04:58:49

mastodon - Link to source

Question for anyone on here using Google Gemini.
I want to extract some text from some war diaries and preferably save it as an HTML file.
I realise Gemini probably will not create an HTML file for download but it did offer to create me HTML code.
I put it into a text editor, saved it as an HTML file but there must be something I'm missing because when I opened the new document in a browser, it just showed me a whole bunch of HTML code instead of my nicely formatted text.
Is there something I'm doing wrong? I'm on the mac.

mastodon - Link to source
Question for anyone on here using Google Gemini.
I want to extract some text from some war diaries and preferably save it as an HTML file.
I realise Gemini probably will not create an HTML file for download but it did offer to create me HTML code.
I put it into a text editor, saved it as an HTML file but there must be something I'm missing because when I opened the new document in a browser, it just showed me a whole bunch of HTML code instead of my nicely formatted text.
Is there something I'm doing wrong? I'm on the mac.
⇧