Instead of chasing the latest gadgets and trends, let's see what we can do with the hundreds of thousands of excellent robots that are already out there. Saving the planet, one programmable brick at a time.
Wrote up some notes on Clay, a fascinating ~2,000 line dependency-free C layout engine project that can be (inadvisably) run via WebAssembly to define the front-end of a website entirely in C simonwillison.net/2024/Dec/21/…
Fascinating project by Nic Barker, who describes Clay like this: > Clay is a flex-box style UI auto layout library in C, with declarative syntax and microsecond performance. His [intro …
> and it's not clear to me what the accessibility implications are.
The accessibility implications are as serious as you might guess. The links aren't properly labeled, there's no semantic markup such as headings, and since there's a div for every line, continuous reading with a screen reader is choppy, that is, it pauses at the end of every physical line.
Doctolib ermöglicht eine User-Enumeration, was bedeutet, dass durch Eingabe einer E-Mail-Adresse überprüft werden kann, ob diese im System registriert ist oder nicht. Das stellt eine Sicherheitslücke dar, da potenzielle Angreifer so gezielt überprüfen können, welche E-Mail-Adressen mit einem Account verknüpft sind. Oder wollt ihr einfach nur wissen, ob der Nachbar dort ein Konto hat!? 🤦♂️
Bei der Todesfahrt von Magdeburg sind mindestens zwei Menschen umgekommen, eines der Opfer ist ein Kleinkind. Der Täter fuhr gezielt durch einen Weihnachtsmarkt. Die Ereignisse des Abends zum Nachlesen.
A game-changing HIV drug was the biggest story of 2024 In what Science called the 'breakthrough of the year', researchers revealed in June that a twice-yearly drug called lenacapavir reduced HIV infections in a trial in Africa to zero—an astonishing 100% efficacy, and the closest thing to a vaccine in four decades of research. buff.ly/3P2zrt7 #ShareGoodNewsToo
Hello applevis, I started to tinker with emacs by reading its documentation and notice somthing very obvious, macos keystrokes are derived from those of emacs. I just have a question.
Following an earlier threat, Gaie Delap is going back to prison as there is no tag that can be used to enforce her non-custodial sentence as a Just Stop Oil protestor.
As has been pointed out already, this is the result of the firm that provides the tags being unable to comply with their legal obligation to provide the means by which she can be monitored:
Result Gaie Delap will spend Christmas in prison.... because profits come before her rights!
Some insights into where Google is going with #AI, especially Project Astra! Using this. it is really amazing and transformational! If you listen long enough, you will hear that people with #disabilities are mentioned as some of the key customers for this technology. youtu.be/ctWfv4WUp2I #accessibility
In our final episode for the year, we explore Project Astra, a research prototype exploring future capabilities of a universal AI assistant that can understa...
the lack of the option to export all emails in one go is the 1 thing that keeps me from using your services. You can't possibly expect customers to manually select & export all of their messages, if they want to migrate away from Tuta in the future.
At present it's a Hotel California situation: "You can check out any time you like, but you can never leave"
A decent bulk export feature would be on my wish list for 2025. It's been on your roadmap for years. Any concrete developments?
I know that Tuta has been planning it for years. But that wasn't my question.
Apparently #Tuta still hasn't set a concrete, near term date for the availability of the *option to export all emails in one go*.
So: Customers can easily import their emails to #Tutamail, but they have to manually select and export all of their emails if they want to migrate away from Tuta to a different #email provider.
Ich will die Funktion in eurer Mail-App im Browser (falls es ein LibreWolf Problem ist, habe ich nichts gesagt.) mehrere Mails auf einmal auszuwählen, wenn ich shift gedrückt halte. Ansonsten alles sehr nice und weiter so. 👍
@fatherpat Creating an inbox rule to automatically put spam e-mails that match a specific pattern into the spam folder doesn't prevent such spam e-mails from reaching me. I would prefer if e-mails matching my spam filter were outright rejected by the server with an error like “address doesn't exist”.
@fatherpat Related: not having to set always 'Always trust sender'. Each time I open tuta, the same message appears for all senders, even those for which I've already trusted. And for the emails which went to spam, setting this should move automatically those emails to the main Inbox and should avoid future messages from the same sender going to spam.
Eu gostaria de ter as opções de imprimir e de exportar e-mails para PDF, no aplicativo Tuta Mail para Android, que utilizo diariamente. English: I wish I had the options to print and export emails to PDF in the Tuta Mail app for Android, which I use daily.
Same as others, and also, and again (I know I've already asked...): - A bookmarks syncing service (social.vivaldi.net/@IGVazquez/…) such as xBrowserSync - A thunderbird bridge - A *.deb for Debian distributions installation.
xBrowserSync: Browser syncing as it should be: secure, anonymous and free!
https://www.xbrowsersync.org
The main developer does not have time to work on it (https://github.com/xbrowsersync/app/issues/425#issuecomment-1266620034).
Spotify is a creepy, and grossly unethical, platform at this point. it rips off musicians and rewards huge companies -- and itself. Apparently what it's doing is legal, however.
This is great reporting, from a new book by Liz Pelly. A compelling read.
@kurtseifried The main problem in my view is that virtually nobody cares about the CWE. No one ever complains about it or asks for it to get change etc. We basically put a finger in the air during 20 seconds when writing the advisory, search around all the available ones and then go with the one we think is the least bad one. And we never go back to reconsider and no one propose anything else.
So they are all fairly arbitrary but "decent" because we try to pick an accurate one.
@profdc9 we do try to make it hard to do wrong, daniel.haxx.se/blog/2023/12/13… - but recall that 60% of the curl security problems were *not* C problems.
You know I spend all my days working on curl and related matters. I also spend a lot of time thinking on the project; like how we do things and how we should do things.
With #Prusa turning from #openness with its new Core One model and other cases, there is a need to discuss alternative models to cultivate and sustain #opensource hardware. Although promising approaches exist already, it might be vital to identify the necessary patterns of #Commoning. Read my thoughts here: blog.opensourceecology.de/en/2…
The introduction of a new, closed Prusa CORE One should come as little surprise, but there could be a remedy A guest article by Paul Jerchel Commons die without…
Extreme, extreme longshot, but does anyone happen to know of a copy of Everybody can Read by Lew Robins? This is a program from the 90s that's designed to teach children to read, and I'm interested in the Eloquence builds contained in it. I've tried searching on the web, but can find nothing on the program aside from a couple articles in The New York Times. Here's one of them. archive.nytimes.com/www.nytime…
The Doctor's deadliest enemies, The Daleks, made their debut on this day 61 years ago in the classic 1963 story The Daleks - Episode 1 - The Dead Planet, and they’re still going strong! Happy Dalek Day! #DoctorWho #daleks
VMware Workstation 17.6.2 Pro is now free for commercial, educational, and personal use! Check out the latest features, bug fixes, and resolved issues.
This is important: another mortgage crisis is emerging, this time mostly in commercial/office estates. Wondered where all the back-to-office nonsense is coming from? It's the landlords—companies that are up to their armpits in subprime offices are trying to fill them. Same with "AI is coming for your jobs", it's to force the serfs back to the office (ideally for less money). atomicpoet.org/objects/9a15e9b…
Another real estate crisis is emerging. Here’s just a cold hard fact: 10% of loans on office buildings are in arrears. Delinquencies on commercial properties spiked last month. And because interest...
Google finally did it. Is it time to switch to Firefox? I believe users should be in control of what they can install and use on their desktop or laptop. There's uBlock Origin Lite, which replaces the original uBlock Origin for Chrome and similar browsers. Give it a try.
From Envision: Our holiday promotion is still running + how a blind artist is using ally for music & gaming + what's coming in 2025 groups.io/g/tech-vi/message/83…
Extermiknit! Celebrate International Dalek Remembrance Day, the anniversary of the first appearance of the Daleks on Doctor Who on December 21, 1963, with free Dalek inspired knitting patterns on my Doctor Who pattern post intheloopknitting.com/doctor-w… #knitting #DoctorWho #Dalek
Fellow Whovians will find knitting patterns here for any age or level of interest. There are meticulously researched knitting patterns for Doctor Who cosplay and fun patterns to show your love of Doctor Who in your daily life.
So, it's time again for my traditional synth Xmas. Where all voices have been produced by various text-to-speech engines. I am working on a new one, but it's not quite ready. So here's the old one for now. Happy holidays everyone: OneDrive: 1drv.ms/u/s!AsrDn-5JMCwmgYFS47… Google drive: drive.google.com/file/d/14xMhj…
Introducing VScan II: A visual perception layer for the blind
Sensitive content
Perceiving space's geometry lets me find interesting things. Perceiving space's vibes lets me discover them.
Travelling the world as a blind individual, I get to see many interesting and beautiful places. However, while I can navigate around quite well using human echolocation and my long cane, I still felt something was missing compared to my sighted times. I can use echolocation to perceive the geometry of the environment, but not so much its vibes, which often lie in subtle details. A silent, empty square feels completely different to a square full of people, music and life. In a similar fashion, a square in the middle of the day is something else to a night square, lit by colorful neons, full of flowers, statues, ancient or modern buildings.
I can discover a monument using echolocation, but there's little I can do to tell what it represents, what color does it have, whether it's old, new, in what condition. I can walk on the street and notice a restaurant with a terrace, but I don't get to see its visuals, decoration or style on my radar. I can observe buildings on a square, but don't get to admire the magnificence of their architecture.
Indeed, visuals are not the only medium for vibes, I can still get a lot from sounds, smells, activity. But it's a significant-one and I was thinking, whether I could make use of the large language models to fill this sensory gap. Before, my travel routine was to search for cool places on Google, and later find them in the environment. But with a good enough visual perception layer, perhaps it may be possible to do it the other way around. To visit a place, and discover interesting things on the walk. And get an even deeper connection with the space.
Thus I designed VScan 0.2 to research this approach, as well as address some other camera-related needs I experience on my adventures.
Introducing VScan II: A visual perception layer for the blind
Sensitive content
Hey, just found this in bookmarks and finally installed it. I couldn't find an option to specify an LLM/api key though. The list to select a model is empty for some reason.
re: Introducing VScan II: A visual perception layer for the blind
Sensitive content
@jonathan859 hi! The way VScan works with model has undergone a deep rewamp synce v0.2.1. I recommend checking out the latest announcement on my profile and reading the initial section of the readme, it describes how the new system works. The TLdR is VScan now has a very flexible model of defining models and providers running these models, so it's not just about entering an API key and choosing a pre-defined model, you need to define a provider, OpenAI for instance, or Anthropic, TogetherAI, your own server etc., you need to define what models is this provider running and under what IDs (VScan will fill this out automatically for you if you use a preset) and finally, in order to use a particular model in your configs, you configure that a particular model, GPT-4O for instance, will be run by the OpenAI provider, which you specified in the previous step. Or, if you mostly plan to use a single provider, you can choose it as the default model provider and only specify providers for models, where you want something else.
I'm posting a unique audio adaptation of Charles Dickens's _A Christmas Carol_ produced by @mcourcel in 2009 or 2010 (I don't remember which year). Every part is voiced by a speech synthesizer. There are a variety of synthesizers with different accents and voice styles, using both formant and concatenative synthesis. There are substantial passages of narration from the original text. I have his permission to repost this. mwcampbell.us/audio/a_synth_Xm…
You probably got this question a hundred times when you first posted the production, but did you make that synth rendition of "We Wish You a Merry Christmas" yourself? If so, which synth is that?
I finally turned off GitHub Copilot yesterday. I’ve been using it for about a year on the ‘free for open-source maintainers’ tier. I was skeptical but didn’t want to dismiss it without a fair trial.
It has cost me more time than it has saved. It lets me type faster, which has been useful when writing tests where I’m testing a variety of permutations of an API to check error handling for all of the conditions.
I can recall three places where it has introduced bugs that took me more time to to debug than the total time saving:
The first was something that initially impressed me. I pasted the prose description of how to communicate with an Ethernet MAC into a comment and then wrote some method prototypes. It autocompleted the bodies. All very plausible looking. Only it managed to flip a bit in the MDIO read and write register commands. MDIO is basically a multiplexing system. You have two device registers exposed, one sets the command (read or write a specific internal register) and the other is the value. It got the read and write the wrong way around, so when I thought I was writing a value, I was actually reading. When I thought I was reading, I was actually seeing the value in the last register I thought I had written. It took two of us over a day to debug this. The fix was simple, but the bug was in the middle of correct-looking code. If I’d manually transcribed the command from the data sheet, I would not have got this wrong because I’d have triple checked it.
Another case it had inverted the condition in an if statement inside an error-handling path. The error handling was a rare case and was asymmetric. Hitting the if case when you wanted the else case was okay but the converse was not. Lots of debugging. I learned from this to read the generated code more carefully, but that increased cognitive load and eliminated most of the benefit. Typing code is not the bottleneck and if I have to think about what I want and then read carefully to check it really is what I want, I am slower.
Most recently, I was writing a simple binary search and insertion-deletion operations for a sorted array. I assumed that this was something that had hundreds of examples in the training data and so would be fine. It had all sorts of corner-case bugs. I eventually gave up fixing them and rewrote the code from scratch.
Last week I did some work on a remote machine where I hadn’t set up Copilot and I felt much more productive. Autocomplete was either correct or not present, so I was spending more time thinking about what to write. I don’t entirely trust this kind of subjective judgement, but it was a data point. Around the same time I wrote some code without clangd set up and that really hurt. It turns out I really rely on AST-aware completion to explore APIs. I had to look up more things in the documentation. Copilot was never good for this because it would just bullshit APIs, so something showing up in autocomplete didn’t mean it was real. This would be improved by using a feedback system to require autocomplete outputs to type check, but then they would take much longer to create (probably at least a 10x increase in LLM compute time) and wouldn’t complete fragments, so I don’t see a good path to being able to do this without tight coupling to the LSP server and possibly not even then.
Yesterday I was writing bits of the CHERIoT Programmers’ Guide and it kept autocompleting text in a different writing style, some of which was obviously plagiarised (when I’m describing precisely how to implement a specific, and not very common, lock type with a futex and the autocomplete is a paragraph of text with a lot of detail, I’m confident you don’t have more than one or two examples of that in the training set). It was distracting and annoying. I wrote much faster after turning it off.
So, after giving it a fair try, I have concluded that it is both a net decrease in productivity and probably an increase in legal liability.
Discussions I am not interested in having:
You are holding it wrong. Using Copilot with this magic config setting / prompt tweak makes it better. At its absolute best, it was a small productivity increase, if it needs more effort to use, that will be offset.
This other LLM is much better. I don’t care. The costs of the bullshitting far outweighed the benefits when it worked, to be better it would have to not bullshit, and that’s not something LLMs can do.
It’s great for boilerplate! No. APIs that require every user to write the same code are broken. Fix them, don’t fill the world with more code using them that will need fixing when the APIs change.
Don’t use LLMs for autocomplete, use them for dialogues about the code. Tried that. It’s worse than a rubber duck, which at least knows to stay silent when it doesn’t know what it’s talking about.
The one place Copilot was vaguely useful was hinting at missing abstractions (if it can autocomplete big chunks then my APIs required too much boilerplate and needed better abstractions). The place I thought it might be useful was spotting inconsistent API names and parameter orders but it was actually very bad at this (presumably because of the way it tokenises identifiers?). With a load of examples with consistent names, it would suggest things that didn't match the convention. After using three APIs that all passed the same parameters in the same order, it would suggest flipping the order for the fourth.
Matt Campbell
in reply to Simon Willison • • •> and it's not clear to me what the accessibility implications are.
The accessibility implications are as serious as you might guess. The links aren't properly labeled, there's no semantic markup such as headings, and since there's a div for every line, continuous reading with a screen reader is choppy, that is, it pauses at the end of every physical line.
Simon Willison
in reply to Matt Campbell • • •