probably_wrong

Leaving aside for a second that this post is clearly an ad.

There are reasons why you can't just say "here's a new wheelchair, have fun suckers!" and that's because a medical device is major part of someone's life and disruptions that would annoy healthy people can be life threatening to them. Worst case scenario you end up in a Second-Sight situation [1].

Let's take a random example: the page says "If you're heading down a steep slope, the powered wheels ease the descent". When the brakes of my bike suddenly stopped working on a steep slope I could just throw myself to the side and brake with my legs. If you're a using this "mobility device" as a wheelchair and this happens to you well, tough luck. You should have bought a wheelchair instead.

Do not "move fast and break things" when the things are actually people.

[1] https://news.ycombinator.com/item?id=30349871

I don't see why not: once a user states their preferences you can use one of many sortings to order the candidates respecting that and use the result for the matching stage. If a user only gives "height" as criteria and gets a list of users sorted by height, the (probably terrible) results are not a result of applying Gale-Shepley wrong.

Otherwise we're in a "no true Scotsman" situation because no one, not even you, can never truly know what your preferences are.

As a quick reminder:

First there was Reddit, which had an API, a mobile version (i.reddit.com) and a desktop one.

Then there was the redesign, after which there was new reddit, old reddit (old.reddit.com), and mobile reddit.

Then they removed mobile reddit.

Then they restricted the API.

Then they made old reddit available for logged-in users only, which they are now restricting even more.

It's obvious that they're slowly boiling the frog until there's only one bloated interface left.

I have been collecting small code snippets that would make for good tattoos and I have to say, some of these ones would probably make a good addition to the collection.

I'm partially more in favor of "red" instead of "hotpink" for CSS, but that's just me.

> I also don't see a problem with statements about making people jobless

I do see a problem with a company loudly announcing that they are going to make people's lives miserable purely for profit. Leaving aside that it goes against OpenAI's stated mission ("to ensure that artificial general intelligence benefits all of humanity"), the disdain for the lives they are intentionally trying to ruin makes it a problem.

And even if you believe that the transition is inevitable, as it is the case with phasing out combustion engines in cars, anyone reasonable would see that the transition is gradual to give people time to adapt. Instead of doing that, OpenAI is burning cash at astonishing rates, polluting the environment, and killing personal computing with the only aim of being the only ones left atop the ruins. I do see a problem with that.

The comment you're replying to is citing almost verbatim [1] Microsoft’s director of Applied Science, Brent Hecht, who called OpenAI's data collection practices "the largest theft of labor in human history" in an internal memo.

[1] https://techcrunch.com/2026/09/17/microsoft-exec-called-ai-s...

On dialogs: I don't think it's fair to say "it's GTK, not Gnome" when a. the behavior started happening with Gnome 3, b. it doesn't happen in Mate, and c. GTK is maintained by the GNOME Project. Behavior, BTW, that (AFAIK) was exclusive until then to MacOS.

On categories: I don't think the Debian people had a big input into how the menu was supposed to look like. I'm willing to bet that it's the default seeing as, again, it's also the default in MacOS.

But on the main topic: I believe that good design should be invisible, and Gnome 3 was the absolute opposite of that. Take Firefox mobile: one day they asked me if I wanted to try the new layout with the navigation bar on the bottom, I said yes, and I never looked back because it made my life SO MUCH easier. In contrast, Gnome's Wikipedia page has an entire section titled "criticism" detailing entire distros that dropped it and successful forks that kept the old behavior. I therefore feel strongly against the idea that it was "well thought out", or more precisely, I think it may he been "well thought out" but not "well tested with the actual users".

I left Gnome 3 for Mate early on and I never looked back, but here are the issues I remember.

First, dialogs could not be dragged to the side. It took me less than one day to realize that, when it comes to naming a file, I need to look at the document to remember, for instance, the name of the institution I'm writing to. If I can't move the Save dialog then I have to cancel, find the name, remember it, and start the save process again, at which point I also lost the folder I had navigated to.

Showing app icons without categories is useless to me because sometimes I don't know what the app is called, only what it does. I know I have a webcam app and I know it should be under Video, but I have no idea what it's called. Gnome's flat hierarchy makes it harder for me to find what I'm looking for.

Removing the individual window menus and the icons for minimizing and maximizing kills the muscle memory I've built over the last 30 years - same reason why I hate that the X button on Mac doesn't really close the program.

But on the topic of Mac, that's for me the main problem: that Gnome 3 wanted to be MacOS and copied their ideas regardless of whether they make sense to their users. I know the things I mentioned above may sound small but those are constant papercuts that remind me every 5 minutes that I didn't use to have so many papercuts. I can't imagine that anyone did a proper usability test before dropping what, to me, were breaking changes.

There were other factors like the attitude of "this is how it is now so take it or leave it", but the whole process led me to realize that I don't want to think about my window manager and that, if I'm thinking about it, then it's getting in my way of doing actual work.

"Just" is doing a lot of heavy lifting there.

Eating is something one does several times a day every day of their lives. It must be done regardless of how broke or in how much of a hurry you are. I applaud anyone who chooses to add this stress to their daily lives voluntarily till the end of their days, but I'm not going to demonize someone for choosing to reject the time-intensive thing in favor of an alternative that's completely normal, cheaper, and faster.

For me it was a major revelation when it was pointed out that, for someone who doesn't want his desktop changed, I did install all sorts of crazy things when I was starting with Linux (Compiz plugins, anyone?). And while Gnome 3 pushed me into Mate, it wasn't because it was different but rather because it was bad. So I'm excited about someone trying something new, particularly someone who realizes that the Desktop is not a tablet.

Here's an idea that I believe I read in HN as "the original purpose of the desktop": add a Desktop Manager that manages which icons are on your desktop and makes it easy to switch.

The desktop is kind of weird, in the sense that it's prime real estate but, at the same time, there's an expectation that you shouldn't put too much stuff there, and then it ends up cluttered anyway. I think it's worth trying a stronger separation between "Desktop" and any other regular folder, emphasizing that the Desktop is where you place links to the files you need right now, but with a mechanism to clear it completely (without losing data) and another to switch which icons you see right now, kind of like browser tab groups. That way, when you're working on a project you place the files you need on the desktop and when you're done you clean it up.

The article disagrees:

> At the time, DoorDash’s campaign contributions in the run-up to the 2025 mayoral election were considered unusually large — it made a single $1 million gift to a super PAC that, at the time, was the largest donation in the race

I was recently job searching and roughly 50% of all jobs ghosted me. Just an automated email and then... silence. And when applying via LinkedIn only ~5% of companies even opened my application.

I can believe that your company has a good reason for keeping the job up, but I also know that some job ads are black holes were no one bothers to do the bare minimum including taking the posting down. I have no idea how to differentiate one from the other.

I challenge you to make a post reach the front page without someone commenting "how about this, but with AI?".

Not a meme but rather news. The FBI has recently updated their hiring guidelines and, among other things, they've removed "bestiality" from the list of reason that automatically disqualify an applicant.

Their argument is that, for instance, people who have been coerced into bestiality against their will may want to join the FBI to prevent other people from suffering the same fate but are automatically disqualified from doing so.

For a tongue-in-cheek version you can check late night shows from last week.

I don't think you're being objective either.

Yes, HuggingFace was audited by a third-party. But said third-party wrote that they had to use unreliable AI in their conclusions (page 26) because they had six days to analyse 1300 (page 70) chains of thought and 70000 messages.

> OpenAI was founded to develop safe AI.

If that were the case, then they would have followed their own report saying not to release GPT-3. Or, if they were following their own founding principles, they wouldn't have stopped releasing models due to (in their own words) the challenging market.

I'm missing a critical part of the explanation.

If "the server has a memory safety bug that allows a malicious client to overflow some buffer and overwrite, for instance, the contents of the cgi_bin_path variable", then why is this a data-only attack? Instill need to overflow a buffer the "traditional" way.

Perhaps this is a good use case for AI? I'm sure it's trivial to write an AI system that reads AI text and replaces AI-generated text with AI-generated, non-AI-looking text.

AI.

From this article [1] published by the European Commission:

> The 3 € fee is applied per declared item, where ‘item’ means each unit of a product that is declared on a separate line. Specifically, the Commission defines "item" as one or more goods in a consignment sharing the same tariff classification, description and, where applicable, origin.

Now, the type of declaration matters [2] - if you're filling an H1 (standard customs) then you pay per item but if you're filling an H6 (postal declaration) or an H7 (Courier or low-value consignment declaration) then you pay per category. So anyone saying that it's always one or the other is, at the very least, oversimplifying.

[1] https://trade.ec.europa.eu/access-to-markets/en/news/eu-appl...

[2] https://www.avalara.com/blog/en/europe/2026/07/eu-customs-h1...

To quote one of my teachers, "you may read the Constitution and think it's all obvious, that it's like writing a law that says people need air to live. The problem is, if you don't explicitly write it down someone may actually try to steal the air".

I didn't need someone to tell me that FIFA is full of ads. But studies like this provide a precise, measurable target to work with.

It is my understanding that you pay per product category. If you get all your electronics in one sending then you pay only 3€ total and not per part.

A major factor is that the companies pushing for this technology are spending significant resources into making sure your consent isn't relevant. From this article [1] about Meta and smart glasses:

> Five years ago, Facebook shut down the facial recognition system for tagging people in photos on its social network, saying it wanted to find “the right balance” for a technology that raises privacy and legal concerns. Now it wants to bring facial recognition back.

> (…) The Silicon Valley company has been conferring since early last year about how to release a feature that carries “safety and privacy risks,” according to an internal document viewed by The New York Times. (…) Meta’s internal memo said the political tumult in the United States was good timing for the feature’s release.

> “We will launch during a dynamic political environment where many civil society groups that we would expect to attack us would have their resources focused on other concerns,” according to the document from Meta’s Reality Labs, which works on hardware including smart glasses.

[1] https://www.nytimes.com/2026/02/13/technology/meta-facial-re... or https://archive.is/65Fze

I'd like to throw an idea to those who know better than I do.

Ever since I started living on my own I've lived in apartments, none of which had a backyard (one of them had a parking lot). None of my friends owns a house either. I'm the perfect customer for this and I'd love to do my own projects (two of which require a lathe!) but I simply don't have the space.

Is this just me? Or is the disappearance of the hobby garage a real phenomenon? A 3D printer is fairly limited in what it allows, but it also doesn't throw sawdust in the center of my living room.

> The Children’s Television Workshop, the maker of Sesame Street, entered into talks with Psygnosis about a Lemmings television show. (…) Television shows, whether aimed at adults or children, are built around memorable characters. (…) So, an edict went down from Psygnosis to DMA to find a way to give the lemmings more individualized personalities.

I wonder if anyone brought up Fraggle Rock [1] during these discussions. The Doozers, a race of tiny creatures whose whole purpose in life is to build structures, feels like a perfect fit for the lemmings.

[1] https://en.wikipedia.org/wiki/Fraggle_Rock

> The thinking guy with the one approach will likely lose a photo competition to one of the accidentally better shots of the person who tried a lot more.

If that were the case then photo competitions would have been upended by now by tourists who accidentally took masterpieces with their phones.

I can take as many thousands of pictures as I want but, without taking the time to critically review them to see what works and what doesn't, I will rarely produce a prize-worthy picture. And then there's the risk that I'll discard it anyway because I don't have the eye to separate the prize-winning picture from the Instagram shot all my contacts are publishing.

First, to get it out of the way: I'm seeing comments here that clearly haven't read the article. I encourage you all to do that.

I feel the article gets it right when suggesting it's the amount of (not always accurate!) data they can throw at you. In an honest discussion there's an assumption that the other person won't straight up lie to me so if someone shows me ten examples for why my argument is wrong I may be inclined to believe them. But if half of those examples are made up, well, that's a different story.

I still think of the commenter here who said "LLMs are a DDOS on free resources" and I feel the comparison works here. If police officers can overwhelm innocent people into confessing, then so can an LLM that "can't be bargained with, can't be reasoned with, doesn't feel pity, or remorse, or fear! And it absolutely will not stop, ever, until you are"... convinced.

x and y are common shorthand for inputs and outputs respectively. Saying "y=LLM(x)" denotes that the input tweets (x) are fed to a black-box function (LLM) that returns a classification (y) for that input. Given that we're talking about binary classifiation it's expected for y to be a value between 0 and 1, which leads to defining LLM(x) as "a function that receives a tweet x as input, uses an LLM, and returns a value in the [0, 1] range representing its classification as ironic or not".

The point about beta (which I think the equation doesn't actually reflect) is just to indicate that this approach risks nothing because, worst case scenario, the weights you assign to the model can simply return the original LLM prediction. Keep in mind that 0*inf=0 and that LLM(x) only returns 0 or 1, so it's not accurate to say that the sigmoid would always return 1.

The function I() is the indicator function [1] which, in this case, returns 1 if the argument is true and 0 otherwise. It's only there to convert booleans to integers because summing booleans is not defined.

[1] https://en.wikipedia.org/wiki/Indicator_function

> How else are they to monetize text outputs?

Sam Altman famously said in 2024 "I kind of think of ads as a last resort for us for a business model." So perhaps we could ask him first what his other plans were and why they didn't work.

The fact that ads are one of two types of content for which people regularly install blockers (with porn coming second) speaks volumes about how useful they really are.

> Because what else can you do?

You can resist.

Upvote quirky projects, call out fake hype, make your disapproval visible, call AI slop what it is.

The term AI has become a liability in some circles, so let that be proof that you don't have to just accept whatever AI companies tell us.

I was wondering this myself so I wrote a script that shows that there are duplicated puzzles if you count a rotated Sudoku as being equivalent. Here's one example:

Puzzle 2 is

  1234
  3412
  2341
  4123

If you rotate it counterclockwise you have

  4213
  3142
  2431
  1324

And if you normalize it, replacing the first row with "1234" (brilliant idea from the post), you get

  1234
  4312
  2143
  3421

Which is listed as puzzle 7. A quick check gives me that puzzles 1, 2, 3, 5, 11 and 12 would be unique under rotation, but I wrote that check in five minutes so there must be bugs somewhere. Also, I performed no check for mirror symmetry whatsoever (update: I did a quick one, same result).

I don't want to miss my chance to say that the post is brilliant and that it convinced me to leave what I was doing to check for rotations. I was nerd sniped in the best way and I take my hat off for the OP.