ManMachine
@max@manmachine.me
Someone used Claude Opus 5.5 over a week to clean-room clone 7 Adobe Creative Cloud apps in Rust.
They’re nowhere close to replacing Adobe, but that’s almost beside the point.
What’s notable is the fact that AI can get you from zero to a recognizable, somewhat functional Photoshop clone in a week, which is still quite wild.
@carnage4life One of the aspects of AI that I don’t really see discussed here on Mastodon is how some individuals and organizations are using AI to make their own in-house replacements for subscription-based software or services. It’s hard to feel sorry for the rent-seeking vendors losing business because of this.
This doesn’t mean the homegrown vibe-coded apps are good or secure, but it’s pretty funny.
@MisuseCase @carnage4life AI discussions on Mastodon tend to focus on the downsides, but LLM development as a way of freeing oneself from rentware is something I've noticed, and appreciated, for a while.
I recall someone else built an accessibility tool—a screen reader—that was a clone of some app that cost like $50/month. I'm fine with that. Most criticisms of AI are entirely legitimate, but I absolutely hate the greed that defines app development today, and it's good to see some pushback.
@max It's true that you have to rent the computing power for building these apps. I have less of a problem with paying for external resources I use, vs. paying for the artificial privilege of using my own computing resources to run software on my own machine.
I think the key difference is that if you stop paying for your LLM plan, you still have the things you built with it. Whereas you could sink thousands into some rentware app and have nothing—maybe not even your files—if you cancel the sub.
@csilverman @max i think this bit is why its so important to use the ai to make an actual old fashioned application you ca build and deploy.
Making tooling that uses the ai live is a massive risk
@max @aliengasmask Exactly. That's the downside, of course: AI is basically like hiring a developer to build you something you couldn't build yourself. If you want bug fixes or upgrades, you have to hire the AI to do it.
I still see that as a benefit over rentware, though. Most apps—especially web-based ones—can last for years before you need to upgrade. So I'd look at it as buying a new version of your app every few years, vs having to pay on a monthly basis just to keep it from being bricked.
@csilverman @max @aliengasmask This also means that if you’re a fairly large enterprise you don’t have the bottleneck of figuring out software licenses for, e.g., Adobe Acrobat, because you’ve made your own PDF editor that can do the job and you can give it to as many people as you want.
I for one hope the Russian plague goes all over the world and kills half the human population. Have a nice day.
@limneticvillains there are rumors right here that there's no plague at all, and it's just a thing that Putin came up with to close the borders and prepare for mobilization. But even if it's real, there are always conspiracies like that, yk
@max Oh also apparently the Russians had already developed a highly antibiotic resistant strain by the 1970s.
@limneticvillains
You saying we're gonna have another pandemic? I still haven't fully recovered from the last one.
Doom with pencil
Rach "pumpkin spice butt girl" [she/they] » 🌐
@YKantRachelRead@social.treehouse.systems
every branch of right-wing pseudo-intellectualism is named something like "super-rational objective humanist altruism" and then you look at what they believe and it's some variant of "corporations should legally be able to murder kids"
falsehoods programmers believe about falsehoods
falsehoods programmers believe about programmers
falsehoods falsehoods believe about falsehoods
When I tried to explain my experience with LLMs to my therapist, I came up with a metaphor of a vegan world where businesses suddenly started to put animal products everywhere and normalise it. Later I noticed that there are many parallels between being anti-LLMs and veganism, e.g. there are different reasons for this, and different harm reducing practices. And then I realised that the metaphor holds because both are ethical discussions about issues on consumption affecting everyone. What I mean is: lots of similarities can be found between LLMs, tobacco, vapes, and I guess smartphones, social media, and so on.
Incidentally, seeing how many metaphors do not hold, it is possible to roughly scope the area of erhics relevant to LLMs. E.g. metaphors about consent and relationship generally don't hold all too well, while metaphors about smoking in public or asbestos in walls of schools and hospitals somewhat does hold.
I guess every metaphor, by necessity, has a limit of applicability, but it is interesting to notice that some of those can be stretched really, really far, making me think the similarities between phenomena might not be entirely coincidental.
@nina_kali_nina how many litres of water for 1g of protein 🤝 how many litres of water to add two numbers together
@neauoire yet another angle: I've seen someone insisting that using existing models is okay, because all or most of the harms are coming from training, and the training costs have been baked in. This makes me think that it is not dissimilar to saying that leather products and animal fats in cosmetics are fine because they're just by-products, with costs of animal lives paid anyways. Except this "AI leather" is also toxic and causes "prAIon disease"
@nina_kali_nina @neauoire I always start to fume whenever I hear this "but, but the animals will be killed anyhow so it doesn't matter" take. *sigh*
Very good analogy. It is somewhat frustrating that we need these but most people seem to have been conditioned into a state where they not even remotely consider ethical behaviour a thing.
@nina_kali_nina me complaining about LLM use at work has been likened to "an annoying vegan" so... definitely tracks
@nina_kali_nina Worth considering.. I live a similar reduction lifestyle and I never considered that it directly factors into my loathe of AI. It's not something you can just negotiate out of someone..
@cuirass_entertainment I guess some people just decide for themselves that it is too hard. That they aren't ready to do make the choice, that it'd require sacrifice, or that it's okay to use because everyone is complicit. Complicated ethical issue :(
@nina_kali_nina @cuirass_entertainment but... it's not hard to not directly use genAI/LLMs
entirely removing any hint of them from your life might be a fair bit more difficult, but there's basically no reason to use one directly and it doesn't do anything better than a person, whereas animal products have their uses (flavouring, micronutrients, flavouring) in a way that doesn't actively deteriorate your cognitive capabilities 
@OctaviaConAmore @cuirass_entertainment whether you can opt out from using LLMs depends on your occupation, sadly. People in academia and software engineering often would have to give up on their careers to avoid LLMs, even if they hate them.
And it is made worse by being put directly into Google Search, Google Translate, default DuckDuckGo, into support bots for anything consumer, into... Everything.
@nina_kali_nina This is pretty much the exact thought process that inspired my short story that did the rounds a year ago https://blog.foxtrotluna.social/theyre-putting-blue-food-coloring-in-everything/
I made it about "blue food coloring", and it was interesting how many people didn't realise it was about AI (and all manufactured consent stuff, really), and instead thought it was about being vegan, or gluten intolerant or something.
@foxtrotluna I thoroughly enjoyed this story when it came out, thought it was definitely about it, internalised and forgotten about it. Thanks for sharing it again
@nina_kali_nina I can’t think of a greater praise than it clearly having enough of an impact to internalize like that ^^
@foxtrotluna @nina_kali_nina @aparrish This is a great story; I hadn't seen it before. I am surprised that "we are a blue-first company" (which was a great inclusion) didn't give it away for everyone. 🤣
As a very long-term vegetarian, I've been feeling for several months now that avoiding AI is feeling more and more like being vegetarian, and I don't mean that in a good way.
@foxtrotluna @nina_kali_nina @aparrish I've never been particularly evangelical about being vegetarian; it'd be pretty clearly better for the world & most people if more people were veg, but also food is deeply tied to psychology, family, culture, and so on.
But AI has none of that! It's like an always-vegan world suddenly introduced a climate-destroying mass cruelty system that worsens your health and the best argument for it is "yeah but have you tried bacon?".
@foxtrotluna @nina_kali_nina @aparrish (On the other hand, this AI crap has sometimes made me wish I was more evangelical about being vegetarian, because good Lord has it been disturbing how utterly alien the idea that you might actually give something up that you enjoy or find value in for ethical reasons is to most people.)
@nina_kali_nina digital veganism is definitely a term that pre-dates the LLM era. It was getting bandied about for people who insisted on open, transparent, efficient tech so as to opt-out of surveillance capitalism. I'm definitely one of these, and have been for decades. LLMs are everything I objected to previously supercharged. So, yeah, I definitely think the metaphor is strong there.
(I usually don't use the phrase digital vegan because I'm not vegan.)
SE Gyges made his long Bluesky thread "AI safety is mostly a sex cult" into an insanely long substack post, then hastily deleted. Fortunately*, it was multiply captured.
https://archive.ph/3B8Bo
original thread: https://skywriter.blue/@segyges.bsky.social/3mvom4b4dn22q
so it's probably important to know this stuff. but it's also important to remember that Gyges is a rationalist who insists he isn't a rationalist, a solid AI doomsday believer who tried for the longest time to hide his power level, and someone who *seriously believes* that you can say slurs against a computer (he compiled the blocklists "AI slur user" and "AI hater" on bluesky).
it's nice to see someone listing all this stuff, but also these are the grains of salt you have to take this writeup with.
anyway, TESCREAL is a cult and Gyges seems to be under the impression he's out of the cult in any meaningful way.
@davidgerard I often think about how - if I hadn't done the reading - I'd think that the rationalist AI sex cult was just some crazy made-up story that unhinged spittle-flecked AI haters use to scare passers-by like a bug-eyed street preacher. It's honestly hard to believe. EXCEPT WE HAVE A STACK OF RECEIPTS. A VERY TALL NOTARIZED STACK OF RECEIPTS.
This remains the stupidest timeline.
@arclight @davidgerard one of the many Worst Things about all this is if you try and describe what’s going on to anyone uninvolved you end up sounding utterly unhinged, even if you try the abridged version of the explanation.
@arclight @davidgerard Though in fairness “it’s a doomsday sex cult that thinks a pile of video game shaders will kill us all, based around a Harry Potter fanfic written by a guy with an eighth grade education who got scared of the computers in 60s pulp science fiction, back before the internet existed” does sound a tiny bit mad.
Running MacOS 27.0.1. I usually wait for the .1 release but was trying unsuccessfully to resolve an issue.
I notice two small things.
1) There is a noticeable delay in waking up after fingerprint unlock, perhaps a full second. There was no detectable delay before.
2) Scrolling in Safari is jumpy now, quantized. Again, new behavior.
Wondering if others see this.
Please don't tell me to run some other OS or web browser, as I already do so in other contexts. I have my reasons to do what I do.
@robpike I haven't made the leap yet, but my boss thought he saw a noticeable battery life improvement. You?
@robpike 27.0.1 on an M1 Mini here. Didn't notice any change in scrolling behavior with Safari. Looks like before. No finger print sensor here.
@robpike I occasionally have three different MacBook’s here, one running 27.0.1 and two others with 26.6 installed. I haven’t noticed any difference in Touch ID response or Safari scrolling. The timing is the same with all of them.
@robpike In the beta a while back they introduced some watch integration and that slowed everything down on unlock. Any chance you have an iwatch thingie?
@robpike Yep, same experience here. Sometimes there is also a noticeable lag when entering text. Not limited to MacBook though; I experience similar hiccups on iPad.
@robpike What I'm taking away from the replies so far is that there is a problem for people called Rob. Apparently names beginning "Ro" that are not actually Rob are also a problem but to a lesser extent.
@robpike I wish programmers would recognize that all software hits a "sweet spot" in its life cycle. After a certain point, changes result in making things _worse_. It's hard to recognize that point, but we have to _stop_ "improving" and "fixing" software after that time.
If the sweet spot is no longer acceptable, it likely means the world has changed around it, and it's time to design a new one to replace it.
The Windows XP startup sound just "played" in my head for no apparent reason🙂↕️
@Aprazeth every now and then I hum passport.mid
Oddly enough for me the most "random" computer music from that era I can distinctly remember is the interlude in Descent. I spent way too much time fiddling with it and sound banks (Yamaha Xg as I recall)
Well that and the Doom2 main menu music. I sometimes randomly and unknowingly tap that one with my hands xD
I couldn't find it, but I remember seeing a comedy skit (in the naughties or teens) where someone was starting up their computer in the library, but the startup tune just kept playing on and on, and they were embarrassed, waiting for it to stop.
I miss when computers were fun. I mean, real unix-like OSes are fun, but can tend to be pretty spartan.
There are no brands, there are only people. Sometimes good people work for a brand and it becomes good, and some people are fooled into thinking the brand is good. Then, without fail, a poorly moulded lump of tofu in a suit takes over and pisses it all down the drain and the brand is bad for a while.
Brands mean nothing. They're just a cosplay costume a bunch of people wear for a while. Who those people are is the only thing that matters.
I do not understand why, but there's an incredible overlap between people who think C is a good language and people who are also just awful
@mjg59 It took me writing my own C compiler to realise just how awful it is. Now I realise that writing a C compiler makes me an awful person, but at least I have a new found respect for Rust's context-free grammar.
@mjg59 it’s a symptom of “the skills that I specifically already invested time in are good and prove I am smart and good, and the skills other people invest time in are dumb and bad. And the fewer people who successfully clear the steep learning barriers I’m actively reinforcing, the smarter and gooder I am!”
that it’s C specifically is just incidental, if history shook out a little differently they’d endlessly defend Fortran instead and group C with the baby languages for babies
Believing that C is a language you just have to deal with is a different situation
@mjg59 Sometimes I ask people what languages they write in, and they tell me they can write in anything; all languages are equivalent.
Reminds me of the interview I conducted where the candidate wrote python as if it were C.
@mjg59 As often is the case, many computer touchers conflate "I have fun using X" with "X is good". I have tremendous fun writing C objectively we should move away from it.
@mjg59 Indeed. I offer some explanations in my 2025 blog post "Rust is indeed woke" https://diziet.dreamwidth.org/19480.html#the-ideology-of-the-hardcore-programmer
@mjg59 Twentieth century end has been too short, too fast. A few years before you were programming with your wisdom, later on you're watching fools programmed with vibes.
@mjg59 I mean, they might be awful, but nothing compared to people finding typescript/node.js good....or even worse; Lisp
Meat proxy is a cool term and all but we’ve had Tech Priest from W40K for decades. It’s all there: wanting to understand the inner workings of technology is heresy that must be stamped out. Everything must be re-produced ad infinitum by performing the correct liturgies. Being obsessed with Mars. Come on
Sacred ancient texts
@heyheymomo are we sure this is not just Sitelen Sitelen?
If you’re ever discouraged and you feel that you can’t draw just be assured that I can’t either but I keep doing it anyway
@heyheymomo I wanted to become an artist/painter as a child. Then I spent over 3 decades believing that I couldn't draw.
Then I did a 'one drawing a day for a month' challenge (like Inktober but in January). After a week and a half, I could not help but conclude that I was wrong.
That realisation hurt...
@heyheymomo I am a firm believer that ANYONE can draw (Or do art in general) with enough dedication. The only exception in this case is maybe people with vision disabilities, but I would only be partially surprised if they were some blind people who found ways to draw (That isn't GenAI of course)
You can't draw YET. In the case of specifically heyheymomo I think you draw well just have a very childlike art style if that makes sense. May not look like very high quality stuff, but it's good.
trying to tune my terminology
| sloppy fashlord: | 22 |
| fashy sloplord: | 50 |
| lordy slopfash: | 14 |
Closed
@thomasfuchs I would suggest that “lordy slopfash” (lordy fashslop?) is the output of fashy sloplord.
Fediverse, the social media system for 40+ year old people who built the world wide web and regret it.
Yep, Windows XP icons were designed and produced on a Mac in FreeHand.
Microsoft was our first paying client back in 1997: we made the icon suite for Outlook Express on Mac OS 9. It was the email counterpart to Internet Explorer 5.
https://news.microsoft.com/source/1998/01/06/microsoft-announces-march-availability-of-office-98-macintosh-edition/
https://techhub.social/@maccatalan/114155754052920748
@chockenberry Some of Dave’s very first concept sketches for Microsoft Whistler (Windows XP) circa 2000. ✍️🤓
@Iconfactory @chockenberry We need more icons like this, and fewer like whatever the heck google is doing.
Perfect! Some decorative spicy pillows to liven up the place...
p » 🔓
@p@hj.9fs.net
p » 🔓
@p@hj.9fs.net