@TehPers@beehaw.org

comments

• The article seems to conveniently hide the rampant corruption between Flock and local governments. Not that Wired has an obligation to include it, of course, but it’s still relevant context as to why everyone hates Flock.

  on Flock Offers Employees Buyouts as Customers Flee in c/technology@beehaw.org · 4d ago

• I love how the Republicans like to talk shit about China, then turn around and propose their own Great Firewall, cameras on every corner, and so on.

  And sure, the Democrats aren’t much better here, but these seem to be pushed more aggressively by the Republicans. Plus, for a party that claims to want less government, it’s very obvious that this goes against their brand.

  on Republican bill would order ISPs, DNS providers, and VPNs to block piracy sites in c/technology@beehaw.org · 5d ago

• This. You’d definitely want the cameras to have only the public key. If they’re not doing that, then their security must be as bad as their morals because they know people are tearing these down everywhere already anyway.

  Of course, not using assymetric encryption makes them more auditable by “third parties” which is an unexpected benefit.

  I wonder if they at least use a different key per camera. This would be less necessary for asymmetric encryption, of course, but they don’t seem to be using it. If all the cameras are using the same key, then you can theoretically use the key these people pulled to access data on every other camera as well (possibly remotely if you can remote in somehow).

  on Hackers Got Inside a Flock Camera. Its Data Shows How the System Really Works in c/technology@beehaw.org · 17 Sep 2026

• FWIW, this seems ripe for data collection and field testing for free. (unless it’s a paid sub, then that’s worse)

  Exploiting abuse survivors to test a surveillance system is a new low —even for government. 🖕🏼

  Did you even read the article?

  Here, I’ll help you:

  The app’s effectiveness depends on courts ordering suspects to wear electronic tags, something that remains relatively rare in South Korean stalking cases.

  Police arrested 29,404 people on stalking charges across 2024 and 2025. Authorities made 1,187 applications for electronic monitoring during that period, but courts approved only 427.

  Your “surveillance system” is court-ordered ankle monitors for known stalkers. Which, sure, is surveillance. You know, for a convicted stalker.

  Whether the ones who were stalked want to know where their stalker is might be a different story, but there’s this cool thing you can do with apps where you turn off notifications and hide the app somewhere you never go. That way they don’t have to see where the stalker is and they still get the benefit of “if that distance falls below 1km, police and probation authorities receive an alert and contact the victim.”

  on South Korea gives 400 stalking victims app to track their stalkers in real time in c/technology@beehaw.org · 15 Sep 2026

• This could be the “new normal” for 2026 and beyond, with monthly totals climbing sharply since mid-year.

  This would require there to constantly be security flaws to patch. Sure, vulnerabilities are inevitable, but steps can be taken proactively to avoid them, and no operating system (most of these vulnerabilities were in Windows) should be introducing an uncontrolled stream of constant vulnerabilities (in many cases against the user’s wishes).

  Anyway, patching security holes is a good thing at least so long as the total surface area of holes actually decreases over time.

  on Microsoft Patches a Record 974 Security Flaws in Biggest Update Ever in c/technology@beehaw.org · 15 Sep 2026

• I don’t fully understand the question, to be honest, but would a purpose-built language like ActionScript qualify? It served only one purpose, and (debatably - AS3 was alright) served it well.

  Otherwise, every general purpose language attempts to serve all its uses at once, which means they rarely specialize heavily into one particular domain. For example, while Rust’s stdlib devs could build a HTTP library, many Rust users don’t need one, good community-maintained libraries exist, and the stdlib devs don’t want to sacrifice stability with something that requires as much churn as a HTTP library needs. It serves their goal of stability really well, but requires extra steps from users who need that.

  on Programming languages are unconvincing (atleast C++, Java and its ilk, Smalltalk, Lisp) in c/programming@beehaw.org · 12 Sep 2026

• Oh hey! Actual man-made horrors beyond comprehension!

  (Sorry, the fly brain simulation stuff felt so tame compared to this field)

  on World-first biological data center packs 16 mn living human neurons in c/technology@beehaw.org · 12 Sep 2026

• This. Unlike the people using actual brains, these simulated brains are, for the most part, just linear algebra and no more a lifeform than a LLM. They lack all of the components of organic life, lack all the chemicals that make a real lifeform actually “feel”, and so on.

  on Google mapped the brain and nervous system of a fruit fly, and modders forced it to play Doom in c/technology@beehaw.org · 11 Sep 2026

• If nobody receives a criminal conviction for illegal hacking, then there was no hacking.

  So pick one. Is it illegal hacking? Or is it marketing? You can’t choose both.

  on OpenAI admits it didn't disclose rogue AI wiki hijacking incident in c/technology@beehaw.org · 7 Sep 2026

• Remember when Claude Mythos/Fable was the end of the world?

  I guess you can’t end the world if the world already ended in 2012.

  on ‘We’re plausibly close to crossing the line’: are warnings of uncontrollable AI coming true? in c/technology@beehaw.org · 5 Sep 2026

• No.

  on ‘We’re plausibly close to crossing the line’: are warnings of uncontrollable AI coming true? in c/technology@beehaw.org · 5 Sep 2026

• It’s mostly just management dictating from the top-down. I’ve heard everything from “it’s our company’s direction” to “investors are only funding AI expenses” and nothing actually grounded on (B2C) customer feedback.

  on Yes, no AI is now a feature - the Document Foundation in c/technology@beehaw.org · 3 Sep 2026

• For software not exempted, the California Digital Age Assurance Act takes effect on 1 January 2027 for requiring age collection at device setup time and age-bracket signals for applications.

  As a reminder, the law doesn’t require age verification, only attestation. It’s a really stupid law that accomplishes nothing for the services that now require age verification and nothing for the people opposed to it. It just makes some OSs (but apparently potentially no longer Android?) ask your age on account creation.

  I still have no clue how this will work for Windows server (and the cursed Mac servers), but honestly it’s on you for using that over a Linux server.

  on Linux (and other FOSS OSs) are exempt from age verification in California in c/foss@beehaw.org · 30 Aug 2026

• Notepad had a “Copilot” button at one point. No clue if it’s still there though since I’m fully on Linux now.

  on Microsoft Paint and Photos Embed Server-Issued GUIDs as Invisible Watermarks in Locally-Generated Images in c/technology@beehaw.org · 30 Aug 2026

• On the one hand, you can’t really believe the US government these days. On the other, the same can be said about the Chinese government.

  You know what would solve this, though? Make the government actually exciting to work for. Get people excited to work there, and you’ll find some talented security people who can lock things down. Instead, all the good positions are offered by companies, and the government isn’t on anyone’s radar especially after Trump/Musk fucked around with it and fired everyone.

  For all the money spent pointlessly on military, you could hire the most talented people in the country to run government services and you wouldn’t even need to sacrifice anything in the military if you just allow them to shop around and buy from the companies that won’t bend them over and make them beg for it. Hell, you could do all that and still provide healthcare to the entire nation.

  on US says Chinese hackers broke into Justice Department, NASA, Federal Reserve, Senate, and more in c/technology@beehaw.org · 28 Aug 2026

• This is such a reach. The best argument Mike can give in favor of face filters is that they’re free speech, apparently, which is dubious (we’ve already decided that under-13s don’t get any social media “speech” at all). Maybe Mike is in favor of removing that caveat too.

  It’s also wrong. Meta has no obligation under the US constitution to protect anyone’s speech, nor does any private individual or business. This is just a case of “free speech absolutism” (aka “freeze peach absolutism”).

  What might (and likely would) be unconstitutional would be a law forbidding children from posting modified pictures of themselves anywhere online.

  on Meta Just Paid Nearly $17 Billion To Make Sure It Gets To Write The Kid Safety Rules For Every Other Social Media Platform in c/technology@beehaw.org · 28 Aug 2026

• Within the US, you should be able to make a 911 call with a phone that has no service as long as it’s able to connect to the cell tower at all.

  I don’t really have any recommendations on how to test this, though. If you have a phone with service, or a phone that once had service but doesn’t anymore, then those would be the devices I’d recommend reaching for in an emergency.

  on Colorado: to test the 911 service, you must first place a normal call to the 911 center. WTF, they don’t think you might be unable to make normal calls (the very reason you need to test 911)? in c/technology@beehaw.org · 28 Aug 2026

• It’s a wall of text that lists too many requirements to keep track of at once.

  Are you looking for hard souls-like games on PC? There’s a lot of good ones out there, but if you find them easy, it might be worth doing a challenge run (no damage, not allowed items or certain mechanics, etc). For example, I had a friend do a no-damage Sekiro run at max difficulty.

  on Question about the 4 hardest videogames among mods combinations and original games either those are not officially pc ported videogames. in c/gaming@beehaw.org · 27 Aug 2026

• The article gives only one example that I can find, and it reeks of “the opposition is threatening us and so we should ignore them”. Their example of the threats against the Marshall City Council members was dropped because the police “did not identify sufficient evidence” that a crime had been committed.

  on Local officials face death threats and gunfire over AI data center projects — More than 500 towns restrict builds as councils shutter public comment in c/technology@beehaw.org · 27 Aug 2026

• For republishing… content that Elon doesn’t own?

  I don’t see the problem.

  on Nitter receives Cease & Desist from X/Twitter in c/foss@beehaw.org · 26 Aug 2026

• Who’s sending them death threats for not persuing data centers projects? Elon Musk?

  More generally, a politician who upsets their voters this much is a failed politician. Even when the populace is split, you can find things that they agree on and work towards that.

  on Local officials face death threats and gunfire over AI data center projects — More than 500 towns restrict builds as councils shutter public comment in c/technology@beehaw.org · 26 Aug 2026

• This. Software is only ever licensed to you, never given.

  What’s important isn’t that it’s licensed (paid software has to be because it’s trivially reproducible), but what the terms of the license state. I’ve read game EULAs that, to my surprise, allowed you to transfer your license once to somebody else (along with instructions on how to do it). I’ve also read shitty EULAs, which is the vast majority of them.

  The biggest scam is when licenses can be unilaterally changed, which is sadly most of them.

  From a game preservation standpoint, the license should be nonrevocable, perpetual, and the game should be DRM-free. Using Steam games as an example, Steam is a DRM, but many games can be played without Steam by simply deleting the Steam API library from the game directory. Those would be playable regardless of whether Steam even exists anymore.

  on Sony Reminds PS Users They Don't Own Digital Games in c/technology@beehaw.org · 24 Aug 2026

• Did you know non American companies are also allowed to have English names?

  This is even funnier when you consider how prevalent English is in NL.

  on ASML stands for Advanced Semiconductor Materials and Laser, and was an American company, so when did it become a French company? in c/technology@beehaw.org · 23 Aug 2026

• Mine lets me adjust follow distance as well, so I don’t actually blame the car. It’s just not the right feature for me when I have no idea if the car 3 lanes over is going to cut across the entire highway to take an exit at the last second. Or for when I’m zipper merging and I have the dipshit who tailgates the car merging in front of me. And so on.

  on One In Five Drivers Switches Off The Safety Tech Meant To Save Their Lives in c/technology@beehaw.org · 23 Aug 2026

• unusable in city, highly situational outside. very annoying when someone merges just in front of you.

  I agree with this, which is why I find myself mostly using it on long stretches of highway.

  sooner or later it will stop working, because it will have trouble reading the lines. people are terrible at monitoring automatic system, that is the lesson we learned from airline cockpits, and i will rather drive my car, than anxiously wait for the moment what the half-assed system stops working. on top of that, my car start obnoxious alarm once it decides i am not holding the steering wheel tight enough, which is about 20 seconds after activation of this function.

  Maybe I’m missing something here, but at least on my car, lane keep assist only corrects you when you’re drifting out of the lane. The system that keeps you centered in the lane is called “active lane keep” which is only enabled for me when I also have cruise control turned on.

  I rarely use active lane keep, but when I do, it’s on a mostly empty highway. Also, I’m not the kind of driver who does shit while driving, so it’s quick at least for me to correct it if it’s wrong (and it occasionally is).

  ehm, are you using your turn indicators?

  When another lane merges into mine? No. The car still has no idea where my lane is though because the line on my right (for the US - I assume left in the UK?) disappears for a short time.

  (i admittedly don’t have personal experience with such system)

  Maybe some experience would do some good then? All it shows me is a number for my speed, a speed limit sign, and sometimes it’ll pick up school zones or no passing zones. It’s tiny and at the very periphery of my vision.

  on One In Five Drivers Switches Off The Safety Tech Meant To Save Their Lives in c/technology@beehaw.org · 23 Aug 2026

• Oh, this makes sense. I don’t think I ever once used sport mode on my old car, and I rarely have a use for it on my new one (basically only if I feel like taking it to a track which is on my list but not anytime soon). Sport mode doesn’t make much sense outside of actual sport cars anyway.

  I’ve seen eco modes, but those seem to be mostly on hybrids I believe? If I could have gotten one, I’d probably have it set to that the whole time for local driving, but hybrids are so popular here that it wasn’t a reasonable option sadly.

  on One In Five Drivers Switches Off The Safety Tech Meant To Save Their Lives in c/technology@beehaw.org · 23 Aug 2026

• By heavy traffic, I’m referring to bumper-to-bumper traffic. I’ve found that I can’t get it to stop anywhere near the car in front of me, always preferring to leave several cars of space. Sure, it’s safer that way, but it’s annoying when the person behind you is laying on the horn while people constantly cut you off.

  Driving myself in those situations is generally just less stressful.

  on One In Five Drivers Switches Off The Safety Tech Meant To Save Their Lives in c/technology@beehaw.org · 23 Aug 2026

• Now that I own a car that has any of these features…

  25.9 percent of owners whose cars have adaptive cruise control never switch it on

  Why? Lack of knowledge? With mine, I use it regularly on the highway outside of heavy traffic. It’s very nice, actually.

  Nearly 18 percent steer clear of lane-keeping assist

  Mine, I keep on for the most part. Theres an easy access button to turn it off when it becomes annoying though, which is only really when dealing with lanes merging.

  21.3 percent disable automatic emergency braking outright

  I couldn’t tell you from memory where this setting is. It has only triggered once for me, which was when parking on a weird hill. I was only moving about half a mph though so whatever, and it let me override it.

  Voice assistants go unused by 17.8 percent.

  Well, yeah. My options are fucking Gemini, Siri (but I have no iPhone), or Alexa. Fuck those all.

  Augmented reality head-up displays fare no better: 17.1 percent of drivers who have one refuse to turn it on, and a further 7 percent don’t know how to operate it.

  I’m sorry, what?

  Having the current speed and relevant traffic signs in the windshield is amazing. It makes it so much easier and less dangerous to check the speed. Sure, you can eyeball it while driving, but there are times where knowing the exact speed is important, like when driving past speed cameras. Also, quick access to the speed limit is incredibly useful when driving in new areas.

  The rest of the listed features are just as confusing to me. You bought a car with massage seats. You don’t use them? Seriously?

  Remote start, I don’t use either, but I haven’t had a use for it yet. To begin with, there’s no benefit to idling the engine, and the weather here has been great so no need to ventilate the car or run the AC/heater.

  45.7 percent of respondents say they regularly use reversing cameras

  At least in the US, every new car has them. So are people going out of their way to ignore them? Or only that many people have cars with these cameras?

  While almost all new cars have them, just 27.3 percent of respondents say they regularly use their vehicle’s driving mode. Those aged 25 to 34 were more likely to use them, at 38.3 percent.

  … as opposed to what? The parking mode? Or is this automatic vs manual?

  on One In Five Drivers Switches Off The Safety Tech Meant To Save Their Lives in c/technology@beehaw.org · 23 Aug 2026

• We’re comparing this to the observable behaviors of humans, though. Humanity engages in unsupervised learning without a scoring function (let alone a differentiable one) and, as a result, can not only train on the fly, but can also find its own ways to self-evaluate its learning. Humans also have hormones which change how the brain works even further.

  This is also all disregarding the ability for the human mind to become faster and more accurate on recall when it has more knowledge to search in a field. LLMs can’t recall very effectively, instead becoming less effective the more information it needs to search.

  on LLMs and Humans are Cognitive Cousins – Sean Carroll podcast w/Dr. Chandra Sripada in c/technology@beehaw.org · 21 Aug 2026

• Llm’s that only have one glyph per token can easily count this

  [Citation needed]

  There’s no difference between counting tokens containing the letter “r” and tokens solely representing the letter “r”. To begin with, this assumes the way we encode words in our brain is by tokenizing each letter, which we pretty much know to be false (we can process words without seeing evry ltr of th wrd and even think about them without it being mentioned).

  yeah we train in bursts because it’s computationally cheaper, this hardly means anything important.

  No, this is not how neural networks work at all.

  We train in bursts because we need a ground truth/scoring function to guide the learning, which means we need a way to guide the model in the right direction. Even with unsupervised learning, the model has a way to determine whether it’s going in the right direction, like with GANs.

  Neural networks are also incapable of synthesizing completely new skills from its own weights. Humans have been doing that for the entirety of human history.

  on LLMs and Humans are Cognitive Cousins – Sean Carroll podcast w/Dr. Chandra Sripada in c/technology@beehaw.org · 20 Aug 2026

• What a dumb question. “Cognition” is an abstract concept that just leads to debate over its definition, except that it’s commonly understood that humans have cognition.

  This entire interview also seems predicated on the belief that neural networks work fundamentally the same as human brains, which is an oversimplification of brains. To begin with, neural networks cannot train themselves at inference time (context windows aren’t training). Also, humanity has barely a concept of how the brain works and regularly learns new things about it. What we do know is that they are more complex than a bunch of connected neurons, and that we have no way of modelling something we don’t yet fully understand.

  So do neural networks have cognition? If they did, then we’ve had cognitive AIs for longer than I’ve been alive.

  The rest of the interview is imaginative fiction based on what they think the model is doing predicated on what Anthropic likes to advertise their models do.

  ---

  As basic evidence, if I tell you that the word “strawberry” has 3 "r"s, you can remember that. You can recall that no matter how many conversations we have in between when I told you that and when I asked you again.

  If I teach you how to count how many "r"s there are in the word “strawberry”, you learn a skill. In the future, if I ask how many "r"s there are in the word “strawberry”, you can count them using your new skill.

  If I teach a LLM how many there are or how to count the letters, it can only use that skill or knowlege if it can effectively search for it. Its ability to execute skills (which, to begin with, isn’t it executing the skills but it asking something else to do them) is diminished the more skills it gains access to, and its “knowlege” (aka long-term storage) becomes diluted when it has access to more of it.

  This is blatantly untrue of humans, who can process knowlege and skills they learn, efficiently and quickly search them, and actually get faster and more accurate as they develop related skills and learn related knowlege.

  on LLMs and Humans are Cognitive Cousins – Sean Carroll podcast w/Dr. Chandra Sripada in c/technology@beehaw.org · 20 Aug 2026

• You:

  So why is more pollution from subject A acceptable, but less from subject B isn’t?

  My response (in response to another comment):

  It being drowned out by the advocacy against AI doesn’t mean those voices disappeared, just that there are louder voices now.

  You, with the nihilistic doomerism take:

  By virtue of even posting, YOU are pushing datacenter demand.

  Me, responding to the subject:

  people are upset about all the evils. Being more or less upset about one or the other is irrelevant. You’re just choosing to cherry pick one thing that everyone who’s upset should ignore.

  What was that about comment history?

  Anyway, in response to this:

  What matters (at least now) is the number of services it provides.

  If you’re looking for things to improve, then you prioritise the places where improvement will have the most effect.

  Anyway, if we want to reduce the environmental argument down as much as you are, then we might as well eliminate nearly all life on earth because living creatures produce emissions and consume water. Point is, your counterargument is complete nonsense and reductionist.

  People can be upset by multiple things at once despite your best attempts to argue otherwise.

  on ChatGPT, LLMs, and Automating the Fun Out of Roleplaying: Don’t let billionaires poison the beautiful creative collaboration at the heart of TTRPGs in c/gaming@beehaw.org · 20 Aug 2026

• Current non-AI datacenters run all of every other service on the internet.

  Current AI datacenters run one class of services.

  This is like comparing totals vs per capita.

  on ChatGPT, LLMs, and Automating the Fun Out of Roleplaying: Don’t let billionaires poison the beautiful creative collaboration at the heart of TTRPGs in c/gaming@beehaw.org · 20 Aug 2026

• you better license it AGPL 3.0 or you’re part of the problem.

  They ignore copyright. License it however you want, but they’ll still train off your content.

  on IP can’t save you from AI in c/technology@beehaw.org · 19 Aug 2026

• By virtue of even posting, YOU are pushing datacenter demand. These comments don’t just live in the ether.

  I write software for a living. There’s an enormous difference between a text-based service that functionally runs as a primitive CRUD and a compute-heavy service like LLM inferencing. The latter runs on servers that individually consume multiple kilowatts while the former can run on a raspberry pi but gains efficiency by sharing a server that consumes maybe 1kW with dozens of other people. Heck, Lemmy doesn’t even need to do video transcoding like all the toxic social media sites do.

  The point being you are MORE upset about the lesser evil, not that one is justified evil and the other isn’t.

  No, people are upset about all the evils. Being more or less upset about one or the other is irrelevant. You’re just choosing to cherry pick one thing that everyone who’s upset should ignore.

  on ChatGPT, LLMs, and Automating the Fun Out of Roleplaying: Don’t let billionaires poison the beautiful creative collaboration at the heart of TTRPGs in c/gaming@beehaw.org · 19 Aug 2026

• Just adding to your comment lol.

  on ChatGPT, LLMs, and Automating the Fun Out of Roleplaying: Don’t let billionaires poison the beautiful creative collaboration at the heart of TTRPGs in c/gaming@beehaw.org · 19 Aug 2026

• Before AI, people had been advocating against the meat industry, and had pushed for reduced water usage in agriculture and homes, which led to stricter rules against lawns in drought-prone areas. It being drowned out by the advocacy against AI doesn’t mean those voices disappeared, just that there are louder voices now.

  I do enjoy the classic “they’re doing bad so I can too” fallacy, though.

  on ChatGPT, LLMs, and Automating the Fun Out of Roleplaying: Don’t let billionaires poison the beautiful creative collaboration at the heart of TTRPGs in c/gaming@beehaw.org · 19 Aug 2026

• With how much shit PBS (and the former CPB) have taken these past two years, it’s shocking they could keep the data in any data center. Also, given that this is in a deep red state, I can’t imagine them getting a lot of support there.

  on PBS station fears losing 50TB of data after being ghosted by cloud storage provider in c/technology@beehaw.org · 15 Aug 2026

• You can block Facebook ads by blocking the Facebook domains. This is what I do, anyway.

  Even if you get through the ads, there’s nothing of value there. At best, you’re maintaining superficial relationships while wading through AI slop and ragebait. If you haven’t actually talked to the person in years, consider whether there’s even a real connection there or if you’re just inflating each other’s egos by pressing a like button every couple weeks.

  on AdGuard promises to keep filtering Facebook ads in c/technology@beehaw.org · 12 Aug 2026

• Sorry, forgot about a few. Sam Altman, Sundar Pichai… Honestly, it’d probably be easier to list the white cishet billionaires in tech than it would be to list the exceptions. There’s tons of them outside of tech, probably the majority tbh, but within tech it’s far too diverse.

  on Top economist warns that the AI math doesn’t make sense: 'Profits are currently being funded by investors rather than earned from customers' in c/technology@beehaw.org · 12 Aug 2026

• Don’t forget about the other big players! Jensen Hwang, Lisa Su, Lip-Bu Tan, Satya Nadella, plus the memory cartel (Samsung, SK Hynix, etc.)

  This particular war is a class war.

  on Top economist warns that the AI math doesn’t make sense: 'Profits are currently being funded by investors rather than earned from customers' in c/technology@beehaw.org · 11 Aug 2026

• We are somewhat similar to contractors and directly help our customers write their code. Usually this means bringing in some kind of starting template, and it’s easier to do this with a permissively licensed template than make them sign another legal agreement.

  They own the code once we’re done, and they don’t need to make it open source (companies naturally don’t like to open source their products). Our templates are open source and well-documented, though.

  on Open-source Licensing 101 - alavi.me in c/foss@beehaw.org · 9 Aug 2026

• We’ve managed to convince our workplace to allow us to use MIT on a few projects. We mainly share that code directly with customers, and it’s easier to do that with permissively licensed code, plus we can make it public for an extra win (both at work and for the community).

  There are definitely good uses for permissive licenses. In the same way that you should consider not using one, you should also consider whether one meets your specific needs.

  on Open-source Licensing 101 - alavi.me in c/foss@beehaw.org · 9 Aug 2026

• “Such speech may be offensive to many, but it is not Meta’s role to police offensiveness,” the company added.

  I don’t really know what else to write here because the quote is already enough. Anyway, if anyone still had any doubts about Meta for some reason, there you go.

  on Meta Caught Paying Nazis to Post on Facebook in c/technology@beehaw.org · 9 Aug 2026