Meta recently rolled out a Live AI feature for its Ray-Ban Meta smart glasses. Unfortunately, it’s hard to know when you should use it — and why you’d want to.
Absolutely zero surprise here. One of the biggest reasons we need real OSS (including data, as much as possible) to win in AI is because the alternatives are extremely grim from not just a speech perspective, but from a simple “whose facts” perspective. mastodon.xyz/@johl/11390104008…
It’s flawed as hell, but I do think there’s correlation between inculcation in the traditional skills around small-d democratic, Western, classical liberal civic values, and participation in FOSS. A software industry that centers those values [definitely not our current industry, nor China’s] is good for all of humanity, in the long run.
Anyone here witnessed/seen hour+ unskippable ads on #YouTube? Yes, you read that right. Supposedly, Google is only showing this to some users it suspects of using an adblocker.
We all know how #Google has become increasingly hostile towards adblocking technologies in the last couple years… but if this is true - and not some sort of bug - geez… doesn’t help they neither confirm or deny it in the news sources I’ve seen.
Ahora que los AI-bros de Silicon Valley están panicando con la salida de #Deepseek y sus capacidades a la par de las de #OpenAI aunque a una fracción del precio, además de ser en buena parte open source, una crítica habitual al modelo es que es chino y, entre otras cosas, te censura cualquier mención a los sucesos de la Plaza de Tiananmén en 1989.
Pues vamos a ponernos empíricos y hacerle la pregunta en local para ver qué hay de verdad. Nos bajamos la versión 14b del modelo y le pedimos un resumen de esos hechos.
Oh, sorpresa. Habla de represión, de censura y de una auténtica masacre.
El modelo no será perfecto y tendrá sus sesgos como todos. Pero criticarlo por ser chino no se sostiene.
After 4 years at AWS, I've left the Rust Platform team.
It's been a privilege working with so many great coworkers, doing everything they can to improve Rust in every way possible. I can't wait to see the results of their projects.
I am proud of my work to improve Rust's developer experience, and I plan to continue doing that.
@Adorable_Sergal Hi there. Tuta's servers remain secured and protected in our data centers. They will remain protected regardless of political changes. At Tuta giving our users the best privacy and security is our goal, and when such forces try to change this, for example Chat Control we actively fight to stop it.
I had not realized just how bad the news of the DeepSeek chatbot had been for Nvidia's stock price. I suppose that placing all your bets on a bullshit technology that nobody wants has its downsides.
FYI Nvidia's stock is still falling. Down 15% now. Considering they're the only company that actually made money from the AI bubble this is very bad news for the entire sector. And good news for humanity.
That being said let's hope that Nvidia is not the Lehman Brothers of 2025 because the last thing we need at this point is another 2008-style financial meltdown.
Hey friends! It’s nearly time for #FOSDEM, the annual free and open source event in Brussels. This weekend, the Mastodon team will be based in building H.
We’ll be raising funds for the project with some of our merch - beautiful limited edition winter mugs, pins, and t-shirts. We’ll also have fun stickers to share for free, so you can show your support for Mastodon everywhere!
(just so you know: we’re a bit short on larger t-shirt sizes, so come early if you want to buy one of those)
Just finished listening to the flourish systems change podcast episode with Dan Hill. Much of the discussion was on #DarkMatter in urban design. Happy that he mentioned the work of @darkmatterlabs.org.
Scariest cable I have that I actually use. It's a USB-C to Thinkpad "adapter" that I bought to power a thinkpad that shipped with a giant 135W brick-of-a-power-supply. This cable does work, but has the tendency to "overload" many USB chargers, causing them to reset. Fun times, but good for traveling so I don't have to lug the brick around with me as well.
@bagder some thinkpad require 135w or even 200w power bricks. I’m not sure about the 135, but the 200w I have predates (2021 model) the usb-c of standard that would allow that amount of power.
With my new PR, you can write "curl --url @file" and curl will download all the URLs in the provided file as if -O was used for each one of them. It can also get the list from stdin if you do "--url @-" in style with how other curl options work.
Tried out the new and popular “Deepseek” LLM with my standard “tell me facts about the author of PCalc” query. At least half were misleading or straight up hallucinations. LLMs are not a suitable technology for looking up facts, and anybody who tells you otherwise is… probably trying to sell you a LLM.
alright I forgot to comment on this, because it was around 2weeks ago, but i'm gonna go ahead and now comment on it. it was thanks to a certain post that I remembered to talk about it. so i'm gonna be doing a interview with humanware in some point in the future. and my tvi (teacher of the visually impaired for more context) basically said, hey don't talk about security. are you kidding me dood? don't talk about security? right, let's ignore the fact that the braille note is like 8fucking versions out of fucking date!!!!! let's also ignoe the fucking fact that this thing could be easy to hack if it gets into the wrong fucking hands. o, right, and let's also ignore the fact that there is literally a critical fucking vulnerability within android8 that the august update fails to fucking patch!!!!!!! forbes.com/sites/daveywinder/2… what's the excuse for not talking to humanware about security. "o it's fine, you've locked down your braille note enough". I get it, i'm basically a security guy, I have Google advanced protection, I have a duress password, I have an antivirus sitting on the braille note, I have keysofts software set to i'm fucking paranoid mode. but really? simpl don't talk about security? you're joking! if you really needed extra features humnware, just make your own OS based on Android, what is it, 16? you're just trying to be cheap so you can get your money from states and organizations around the world...
Google has confirmed a critical security threat for millions of Android 8, 9 and 10 users that could permanently kibosh your smartphone with a single malicious message
@fireborn Yeah you really need to go into the interview knowing what type of questions they will ask you. I suspect its not due to any technical knowledge you may have on security.
As some people already mentioned here or here, Copilot purposely stops working on code that contains hardcoded banned words from Github, such as gender or sex. I am labelling this as a bug because ...
Interestingly, one person captured by big tech found that Microsoft was perfectly in their rights to POST to you "since we could have no proof a user wasn't behind initiating that POST". Another reader wanted Microsoft to be prosecuted for hacking. Others chimed in how this explained their subscribe/unsubscribe links being weird lately: berthub.eu/articles/posts/shif…
tl;dr: Microsoft and other email security scanners will visit the links in email you transmit, and run the JavaScript in those links, including calls that lead to POSTs going out. This used to be unacceptable, since POSTs have side effects.
As the #BambuLab users are slowly waking up to what blog.bambulab.com/firmware-upd… means, it's a good time to remind everyone that the writing has been on the wall from day one.
It's not a "told you so", mind. It's a "this will continue and get worse". #Bambu will continue to tighten the ratchet.
#Prusa is far from perfect, but they deliver reliable #3dprinting based on open source software at an acceptable cost.
And if you missed it: they allow you to maintain warranty with custom firmware again.
As expected, #BambuLab released the standard corporate press release
1. All those things we just invented as the criticism we didn't do 2. Sure, you can use "developer mode" and lose our support 3. We tried to work with others, and it's their fault they didn't do it fast enough and by our rules 4. Everyone else is at fault for this misunderstanding though we might have contributed in the slightest of ways 5. "I am sorry that you are upset"
Someone on #Reddit complained about #BambuLab changing their warranty after they bought something and refusing service based on the new "terms". They used @internetarchive to prove that the change was made.
Das ist ein neuer alternativer Android-App-Store, der von sich behauptet, auf Privacy und Security fokussiert zu sein. Er wird auch über den @GrapheneOS Appstore angeboten.
Bei genauerem Hinsehen finden sich allerdings Apps, die Werbe- und Analysetracking enthalten, wie ich nach kurzer Prüfung feststellte.
hey, on the topic of the curl mailing list: is it available via a public-inbox archive somewhere? I had a quick peek but didn't immediately see a link anywhere.
edit: just to be clear, this is only an inquiry, not a feature request
Newswire: Tek Talk meets 1/27/25 at 8pm Eastern and welcomes Jeff Bishop to discuss new Outlook and its replacement of Windows Mail groups.io/g/tech-vi/message/85…
Mit der Spionage-App mSpy überwachen Menschen heimlich die Handys ihrer Partner:innen. Das ist illegal. Jetzt gibt ein Leak Einblick in die Kommunikation mit Kund:innen. Die zeigt, wie skrupellos die Täter:innen ihre engsten Bezugspersonen ins Visier nehmen – auch in Deutschland. #mSpyLeak
Testuji #Phanpy klienta pro Mastodon. Adresa je phanpy.cz/, Phanpy díky @Archosovy je v češtině. Moc pěkně vypadá, tak kdo ještě nezkusil, tomu doporučuji určitě vyzkoušet.
Es gibt darin einen wichtigen Aspekt, der oftmals nicht ganz so klar ist, aber daher will ich das noch mal kurz erklären, warum die Wurzel des Problems mit digitalen Großprojekten in Deutschland oftmals politisch ist. Ganz besonders bei der #ePA. Ein Thread am Morgen. 🧵
Herzlichen Dank, liebe @bkastl, für diese superverständliche Erläuterung eines der politischen Grundfehler bei der #ePA. Ich bin ja auch schon lange unterwegs, um zu erläutern, wie sich politische Entscheidungsfindung ändern muss, um digital sauber umsetzbare und grundrechtlich haltbare Beschlüsse zu fassen
Der gute @lacrosse@social.tchncs.de hat gestern bei @kuketzblog@social.tchncs.de einen sehr umfassenden Artikel zur #ePA geschrieben, der viel sehr genau trifft.
https://www.kuketz-blog.
@usul I have no fixed location this year (as wolfSSL has no booth this time). I mean to try to make sure me and my friends can hand them out in the cafeteria area most of the time. If not, ping me during the event and I can provide a place-of-the-moment.
Tak se zdá, že se po cca. 2,5 letech běhu své solo instance Mastodonu dostávám na limit databáze. Zkusím zapnout automatické mazání příspěvků starších jak 2 roky, co to udělá. #Mastodon #mastohost
@Michal Špondr Mně se moje Friendica (MySQL) dostala na 1,5 GB za půl roku. Jsem na svém železe, takže tam mám ještě dalších 200 GB. A můžu přidat, kdybych chtěl, ale nechci. Mám rád malé věci...
Twenty-six years ago on this day when we shipped #curl 5.5.1, the changelog was easier to gather than what they usually are today: curl.se/ch/5.5.1.html
Podobno jest boom na podrabiane dyplomy ukończenia wyższych uczelni. Młode pokolenie dochodzi do wniosku, że po co męczyć się kilka lat, skoro dyplom da się kupić, wygląda jak prawdziwy.
A pracodawca nie jest w stanie zadzwonić do uczelni i sprawdzić, czy dana osoba ją ukończyła, bo RODO...
Rozwiązanie byłoby proste: powszechna baza informacji o absolwentach uczelni wyższych, najlepiej z publikowanymi w otwartym dostępie pracami licencjackimi i magisterskimi. Żeby można było sprawdzić, co pan absolwent czy pani absolwentka rzeczywiście robili i czy pracy nie napisał im Chat GPT, albo Google. Ale przypuszczam, że wyszłoby przy tej okazji tyle szkieletów w szafie, że klasa polityczna na to nigdy nie pozwoli.
Wystarczyły by podpisy cyfrowe na PDFach dyplomów (my w Polsce już mamy do tego infrastrukturę, wiele elektronicznych dokumentów urzędowych jest teraz autentykowanych w ten sposób).
Alternatywą jest jakiś QR kod i appka do weryfikacji, à la eDowód.
Scholz zu Holocaust-Gedenktag: "Müssen die Erinnerung hochhalten"
Kanzler Scholz mahnt, das Erinnern der jungen Generation an den Holocaust zu fördern. Zusammen mit anderen Spitzenpolitikern reist er wegen des 80. Jahrestags der Auschwitz-Befreiung nach Polen.
Kanzler Scholz mahnt, das Erinnern der jungen Generation an den Holocaust zu fördern. Er und andere Spitzenpolitiker reisen heute wegen des 80. Jahrestags der Befreiung des NS-Konzentrationslagers Auschwitz nach Polen.
For anyone who does have the 22 kHZ Eloquence already or the Tiflotecnia drivers, I uploaded my new files to eurpod.com/upsampled_drivers.z… This contains two files. _ibmeci.py is the IBMTTS modified to upsample 22K to 44K using Audioop. _init.py goes to TifloTecnia's Vocalizer, inside synth drivers and the folder name of the driver. For Eloquence 6.6, driver files are up at eurpod.com/eloq66.zip - people will need to manually run regsvr32.exe ibmcc.dll in their command prompts, and potentially preinstall a ViaVoice5 package first if registering the DLL still does not make it speak. (datajake.braillescreen.net/tts…) It's risky though, when it's broken, the only thing that will save you is Narrator to click through the "entry point not found" dialogs. That's how you know those DLLs are not registered right on your system. Either way, you can hate the upsampled voices, that's OK, it's purely optional and has made my own life better and that's what mattered with this project.
@JamminJerry oh, if the sample rate combobox within the voice settings is at 11K for it, it won't change anything, only at the 22K one. And sadly most people's is at 11K since the 22K one is so hard to get running, so you probably wouldn't notice a difference in that case.
well hell! ok, well I will scrap that idea then, as I bet mine is not at the 22K version as I didn't make any changes. I wonder why the 22K one is so hard to get going?.
@JamminJerry it was introduced too late in the cycle of Eloquence, I think. Also the fact that Eloquence technically had both an IBMTTS ViaVoice version and an ETI-Eloquent version made at the same time, and only the pure IBMTTS form of it had the 22K setting - the one that is a part of JAWS and thus most other versions of the NVDA Eloquence driver use the one made by Eloquent Technologies.
I'm genuinely not folowing (and curious) about the desire to upsample here. Windows already has to upsample unless you literally force all system audio to 22khz. In practical terms, that means Windows normally upsamples to 48khz. So why add the additional upsampling? Are you deliberately trying to trigger aliasing artefacts?
@jcsteh So in short, the Windows mixer is resampling everything anyway, but if you love how certain TTS engines sound when you do a specific upsampling step, you’re not “pointlessly redoing” Windows’ job—it's creating a different audio texture that Windows alone isn’t going to provide. Windows’ built-in mixing and resampling aim to be “transparent,” preserving the original audio as much as possible. It is designed to blend multiple audio sources gracefully—it’s not really trying to color or brighten the TTS. Believe it or not, some folks like the subtle hiss or crispness from naive upsampling. It can make speech feel brighter or easier to parse, even if it’s technically introducing aliasing. If you leave it to Windows alone, you might miss out on those “colorations.” You have more control over final frequencies, filtering, or smoothing, rather than letting Windows do whatever it wants at the last step. Since The TTS engine itself is generating fewer samples per second. It’s basically missing upper frequency content, and even if Windows (or myself) upsample it to 48 kHz, you can’t magically recreate all the higher frequencies that were never in the original signal, that I do recognize. But at least we have finer control over the final audio to have the illusion of treble.
@rommix0 Cool. I have ViaVoice 6.7.3.2 with the NVDA add-on by David CN. I replaced _ibmeci.py in the synthDrivers folder with yours, then restarted NVDA, but I haven't noticed any difference. Do I need to change any option in NVDA settings, or perhaps my laptop speakers are too bad? Also, I haven't found 44 khz amongst the bit rates in the options. There are only 8, 11, and 22 khz.
This project is aimed at developing and maintaining the NVDA IBMTTS driver. IBMTTS is a synthesizer similar to Eloquence. Please send your ideas and contributions here! - davidacm/NVDA-IBMTTS-Driver
@TomGrant91 @rommix0 As for ViaVoice, there used to be an FTP server at IBM where you could download it. I tried to access it, but it seems offline, or maybe I don't recall the address correctly. I believe it was ftp.ibm.com
@clv0 @TomGrant91 @rommix0 I wonder if the 6.7 sounds better than the 6.6 in terms of the 22K voice, maybe. I do know that at the 7.0 I tried, it gets very aliased and hissy sounding, so they broke that 22K behavior in some of the drivers, and that's really why I've stuck with 6.6, it has the best quality for the 22K without any extra aliased high-pitched sounds there. Even when I upsample it, the only noises that sound a bit off are the various pops eloquence naturally makes when certain words follow each other, but otherwise it can keep the 44K resample without making it too hissy.
@TomGrant91 @rommix0 I haven't tried 6.6. In any case, I can try uploading 6.7 to some place if you want. the only caveat is I don't recall whether I've downloaded all of languages at the time.
oh yeah, my eloq66 zip file on that Eurpod link (eurpod.com/eloq66.zip) has an ibmeci66.dll file in there. Looks like my ibmeci.dll and ibmcc.dll files are the 6.7.3.2 one in that package, too. So that's the bad one. That's right, I couldn't get the V7.0 dll file to load into the IBMTTS driver, it told me it's not a valid IBMECI.dll file even though it is, so I could never test how much better 22K is in 7.0 DLL. It's 6.7 I hated though for aliasing, now I remember, and that's why the zip has both: 6.6 which is not so hissy and extra-high pitched like 6.7 is. I do wish I could hear or try 7.0, but beyond just using it as the engine itself, there's no way to get it in NVDA.
![Fire Lord Ozai lies on the ground, defeated with Aang, The Last Airbender, stands before him.
Ozai: [my knees] Wh- What did you do to me?
Aang: [30th birthday] I took away your bending.](https://fedi.ml/photo/preview/640/631807 "Fire Lord Ozai lies on the ground, defeated with Aang, The Last Airbender, stands before him.
Ozai: [my knees] Wh- What did you do to me?
Aang: [30th birthday] I took away your bending.")
Luis Villa
in reply to Luis Villa • • •Hubert Figuière
in reply to Luis Villa • • •