Generative AI summer 2026: where do you stand?

I'm some mix of these myself:

I dislike or distrust gen AI but it draws me in
I do not like gen AI in general but I use it because I sometimes find it useful or fun or interesting
I like gen AI in general and find it fun/useful/interesting
I could like gen AI as a tech but the environmental, economic, political etc issues are too big

so well done me for creating a poll in which my own opinion is not really represented in any clear way.

I messaged some friends just this morning about how much I hate the look of A.I.-produced promos for charity events, businesses, etc. We're inundated with them in a town FB group I help moderate, and they all look the same.

https://i.postimg.cc/ZnpD5Tjx/pizza.jpg https://i.postimg.cc/DyQMVw8N/fair.jpg https://i.postimg.cc/rpsntGQs/yard.jpg

As busy as the claustrophobic '70s movie posters, but blander and much uglier.

Don't remember what I voted last time. But I'd say a mix of these for me:

-I viscerally hate gen AI. I will never use it
-I could like gen AI as a tech but the environmental, economic, political etc issues are too big
-I am just sceptical of gen AI - it does not seem to work well for most tasks I care about

In conversations with people who use it and like it, I usually say that I don't find it useful and that, as someone who does creative/knowledge work, it feels like anathema to use it. What makes the work I do rewarding, after all, is the human brainpower I put into it and the pride I feel as a result.

In thinking more broadly about it, I am depressed by the homogeneous aesthetics of AI-generated content and worry a lot about the social impact of people's overreliance on AI (for instance, students who are unable to think critically, the increase in the spread of misinformation, etc.). I am also concerned about AI putting people out of work, though I am not immediately concerned about my own job. I tend not to focus on the environmental aspect too much because it feels a bit abstract and I feel like I have enough reasons to be opposed to it already. All of this adds up to something like a visceral dislike or at the very least a distrust.

I lived happily without it for nearly 50 years and despite all of the hype it’s not useful to me in my work. The only thing my colleagues use it for at my work is making funny pictures to razz each other. If it disappeared tomorrow no productive work would be lost and life would go on.
The cartoony aesthetic is what really depresses me. The job loss aspect has been way overhyped but it seems like graphic designers are doomed.

There are things that are called AI that are useful to me, but they are not things that are called generative AI. (I personally hate calling any of it AI, although obviously that's not a phrase with a fixed meaning.) I don't remember if I voted in the 2025 poll, but I'm definitely on the "viscerally hate" end of the spectrum in 2026. Literally visceral, I get a real disgust reflex when I see AI imagery or read AI text. And don't get me started on AI music.

I'm still waiting for my settlement check from Anthropic, but even that won't make me feel any more positive about it all.

Can I ask a general A.I. question here, or is there a better thread?

I'd recommend this thread: Artificial intelligence still has some way to go

Mixture of 'viscerally hate' and 'could like except for all the issues'. Voted for the latter as there was a time when I did like it as a stupid toy that created glitchy monstrosities. The issues are pretty fucking big, though.

Visceral hate. The AI generated food images I keep seeing are giving me adult onset trypophobia. How anyone can look at them and find it non-repulsive is baffling to me.

I don't use the consumer version of it and think it's generally bad for the world. I do use it for certain things at work (and work on applications of it, now), and think it actually has positive uses for health care and research.

The development of AI is in the hands of people who can be trusted to wreck every aspect of human society if it enriches them. Whatever AI's current capacity and future potential for promoting human good might be, the blind avarice of AI's owners and investors ensures that its good aspects will be swallowed whole and used to grow and nourish its capacity for evil. We are fucked.

I viscerally hate gen AI. I use it sometimes because I am a lazy hypocrite with poor impulse control.

I think the technology has the potential to do lots of good and interesting things, even in the realm of the arts, but the worst versions of this seem to have won out at the moment. The situation around AI is probably not going to improve unless smart regulations and controls are put into place to help guide its development. The current free-for-all is a mess and encourages all the worst and laziest behaviors while putting people's livelihoods at risk.

a) I am just sceptical of gen AI - it does not seem to work well for most tasks I care about
b) I viscerally hate gen AI. I will never use it
c) I could like gen AI as a tech but the environmental, economic, political etc issues are too big

I'm a mix of a and b, tempered by just a touch of c. Not sure which to vote for.

visceral, baby. now and forever.

I do not like gen AI in general but I use it because I sometimes find it useful or fun or interesting

Text: I've been using gemini occasionally in the last year -- and I don't know how people take these chatbots seriously. If you want to know something that it can't google, it spits out a lie. Sometimes it will give an answer that sounds like it could be right, but then you look into the details and all of the details are wrong.
Image/Video gen: what they can do convincingly has such limited range that I consider it more of a party trick than something useful.
Audio: Suno is 100% slop. I can recognize it instantly.

Also, gen AI has ruined websites and searching. I'd say 90%+ of google search results are AI generated in some form.

Even to spell out my and the university's AI policy in my syllabi taints me.

I appreciate Brian May's approach:

https://www.instagram.com/p/Db46RgQM7Kw/

i hate it but i use it occasionally for c0ding tasks, no full on vibe coding i dont even know how to do the agents tho im sure its not hard to figure out thats the whole point right its easy, its funny when people are like learn ai or get left behind learn what how to talk to the chat bot yeah thats not too hard pretty sure i could pick that up instantly

i did vote I viscerally hate gen AI. I will never use it tho lol

I have friends who use it for coding, that seems like kind of a special use case because it's one of the main things these guys built it to do, right? The first thing they wanted to do was make their own jobs easier. I feel like it's closer to being a specialized tool for coding than a specialized tool for writing or making art. Not perfect, obviously — one of my tech friends says his quality control job has gone from fixing human errors to fixing AI errors.

hate, maybe not viscerally but i won't touch any video game, movie, etc that is confirmed to use it in a creative capacity. i do look at the example code it generates in google searches when I'm looking up a function / programming syntax / whatever at work

I'm mostly indifferent in my job, and we see examples where it could be helpful, but currently trying to steer away from it - not so much from hating AI, but from the boring ass meetings that will result from consultants trying to assess it and 'find solutions'

Yes, AI could probably soft-boil an egg but I don't really want to extend the energy to find out

Art/Music/Literature/filmmaking? It fucking sucks and is utterly without merit, full-stop

quickly surfacing example code buried in documentation i really think it's the only thing it's actually good for. and you could probably develop a specialized search engine for this instead that doesn't destroy the world

regarding coding. if you understand what it's doing and you're just using it for the boring, uncreative parts, I think it's fine. if you're saying "make a website about x" and it makes the generic cheesecake factory website filled with slop, you are a bad person.

extraction is good and useful. text summarization seems decent but you still need to check everything.

i viscerally hate it but occasionally i find it interstng

Things I viscerally hate: the AI industry, its under-regulation, slop, all the dumb ways people use AI, all the ways AI makes people dumber, the techno-utopian and/or managerial rhetoric around AI, the way it seeps into more aspects of life, aggressive data centre development, etc.

Things I dislike but occasionally find useful: Claude, posing questions to google.

I’m currently building a bike, and have found these sources useful for figuring out what parts to get and where they’re available. I’m not confident I could do that by reading 100 reddit threads or talking to people at shops, since it can crunch the specifics of many small details and determine their compatibility relative to my needs (that said, I’m also reading actual books to learn about this). I’ve caught a lot of mistakes and conflicting information, which has unsurprisingly not made me more skeptical of everything else suggested.

I think this is the first instance I’ve used AI at length without feeling a little guilty about it, I don’t have the energy to keep clinging to visceral hate or to keep trying to magically wish it away.

Things that are *shrug*: dull coding stuff that’s above my skill level

I don’t include Google search results in stuff I hate, even though they are using a form of AI. That’s fine to me, that’s just categorizing and sorting things, it’s what Google’s always done anyway in one form or another.

I mostly don’t like or rely on the Gemini search summaries at the top of the results, but sometimes they will have useful links.

But also I’m pretty good at searching for stuff online myself, it’s been part of my job for years. I have ways I like to search and sources I rely on more than others, so I prefer that to asking Claude to do it.

I posted three of those awful, formulaic generative ads/announcements yesterday. Here's another one from today--looks the same, but the best part is the box in the top left, next to Nixon.

https://i.postimg.cc/bwk7HM5s/nixon.jpg

I pretty much hate the idea of gen AI and have never used it.

I scroll past the AI results when using Google.

I do use an AI site to transcribe interviews for me. It never generates anything that I can use without going back over it and correcting things, but it does give me a fast first draft.

Part of me wants to try and use a gen AI program "wrong," give it a bunch of really shitty prompts to generate a really surreal/dumb/annoying piece of writing somewhere between a game of exquisite corpse and a Burroughsian cut-up, as an art project, but I wouldn't even know how to get started, so I'm unlikely to do so.

However much I hated this stupid bullshit, which has now become an integral part of how many of my colleagues do their jobs, it’s been recently completely surpassed by an even more consuming disgust and fury at creepy surveillance glasses (now that those have apparently moved into the mainstream among misogynist teenage boys with a mean streak)

Yeah the glasses are very much a "pull them off the wearer's face, snap them in half, throw the halves in opposite directions" thing.

I think it's stupid and it's strategically suicidal for people to enmesh anything into their lives that's so obviously anti-human. It just doesn't interest me, it doesn't help me, it will make my life better when it does my fucking laundry. Until then gtfo.

The thing that I'm going to be forced to do for work is like...we create a database of our curriculum and documentation and then...make an...agent? That people can ask questions, and it looks through all the materials and finds an answer? I truly do not understand how that isn't just a search...engine...but I don't really understand these things.

The recent illustration I heard on some podcast was:

#1: It seems that AI can be better than human eyes at finding certain kinds of evidence of cancer in some radiology images.
#2: The highest standard of care would be to have an AI that scans images IN ADDITION TO humans, and it tags things for humans to look at more closely. This would require employing both AI and people.
#3: But we won't get that. (That was the podcast; the rest of this crashout is all mine.) What we'll get is AI in place of people, whatever is most inhumane, and people will die because there aren't radiologists looking at their images anymore, and medical practitioners will be out of work, and money will flow from our hands to insurers and technology companies and the private equity that owns hospital systems.

The problem isn't really whatever they're calling AI, but I can still refuse to entertain it.

otm there is/will be a lot of pressure in corporate settings to offload human tasks to AI, which will mostly not do the tasks nearly as well as humans but may do them well enough to not crash the companies. And that's fine with corporations, who are much more interested in not paying humans than they are in making good products or providing good service. They've already been doing that even without AI, but now they can do it and pretend they're not even making any tradeoffs.

Looking back at my answers last year and not much change.

Still feel the lack of returns -- it's not good enough to get rid of enough jobs -- will lead to the bubble bursting.

What is new is how this will intersect with the general state of the economy, which, due to price of oil due to Iran war etc, may speed up that outcome.

But the the tech should remain. It will be an aid to faster task completion. Probably cheaper, more efficient, less environmentally taxing. If it turns out to be that then it has a future.

Feels like we are reaching such a bunch of tipping point re: climate change that will probably put a hard stop on the damage through data centres and the like. Plus ofc that people are hating on that stuff.

In terms of my job is doing a fair amount on a couple of tasks that would take so much longer.

But I have been working weekends lately, still (I wouldn't even be posting rn so it's not nothing lol).

It's not solving the fundamentals, it's like the world of work has some structural resistance to putting AI at it's centre. For the good, because it's not a proper solution that works consistently well at enough things.

But to get the returns it needs to be a lot more than it is.

Visceral hate here, to the point that I have pointedly stopped using the few relatively benign AI-assisted tools I previously used to make life slightly easier (Otter transcriptions, stem separation plugins, etc). Pandora's box and all that.

Same answer as 2025, still useful.

thanks for stopping by

@Paul Ponzi: Otter transcriptions are not generative AI, tho. I don't love that there are generative tools within Otter (transcription summaries, e.g.), but I just ignore these.

I use it exclusively for coding and will not complain if it no longer exists someday.

the only time I used gen ai was when there was a trend of giving midjourney funny commands, I did a couple of those. they were boring and sad and that was it for me & gen ai

I stick to Twink transcriptions to be safe.

@Paul Ponzi: Otter transcriptions are not generative AI, tho. I don't love that there are generative tools within Otter (transcription summaries, e.g.), but I just ignore these.

Yeah I don't care about the summaries, I'm generally giving it recordings of things where I was present so I know what happened. When I look at them, they are generally a mix of mostly right and somewhat confused. I would absolutely not rely on them, and if people are transcribing recordings of conversations that they're weren't present at they should definitely not just rely on the summaries.

With that caveat, I do <3 Otter as a tool, it's awesome.

A long winter for many

NEW: Graduate jobs have dropped by almost 50% in just a year as employers replace entry-level roles with AI

But I still think this reflects the state of the economy, as much as some productivity gains with AI

yeah even the head of Uber just said outloud that AI isn’t providing “return on investment”

And if there's anybody that knows about lack of ROI, it's the head of fucking Uber.

Looking at the thread title, praying no one gets the bright idea to start naming the Generative A.I. Song of the Summer.

Oh my god I heard the worst...pop country??? song at the gym the other day and I had this conviction that it must be AI. Unfortunately for that conclusion, it seems that Kane Brown is a real person.

Kane Browns can use Suno too

pretty depressing how fast this shit has spread into everything, feels like most small businesses are using AI logos now while events are advertised with AI flyers and restaurants now have AI generated photos on the wall, there's a slice of creativity in the world that is now quickly getting completely lost. on the other hand this shit has convinced me of the existence of the human soul, something about seeing a person in a commercial who you suspect may not be a real person sits really badly with me on a spiritual level. it's not like photoshops which feel like a magic trick or the CGI of 5+ years ago which never actually looked real. I think human beings as a whole will reject this stuff the way some animals reject types of processed food. I can totally foresee a study showing advertisements suspected of using AI are completely ineffective.

feels like most small businesses are using AI logos now while events are advertised with AI flyers and restaurants now have AI generated photos on the wall

Adobe subscription prices and getting pretty good at locking down the DRM don’t help IMO. Junior can’t do a quick Illustrator job for mom and dad’s taco shop anymore. It was bad design but human.

It's becoming really hard for me to discern AI slop from regular slop, a factor that AI proponents should consider. If the best AI can do for you is generate something that looks like shit or looks weird or off, maybe there's no appreciable gain from using it. Or maybe you should try harder. The message I get from someone or something all but declaring a minimum effort is good enough is not an enticing one.

What upsets us local film writers is how quickly the indie film houses have replaced interns with AI; they use generative AI for ad copy, film summaries, even the goddamn Dolly Parton obit.

Yeah to milo's point we use AI images at my work and it's because we would need our own photo studio, a professional photographer, and a graphic designer on call at all times to do the equivalent amount of work, and we don't have access to the funding or the tools to do any of those things. We can't even get Photoshop, no one in my whole company has design software capabilities. (This is so weird to me after working in publishing for 15 years when like 1/3 of the office is designers with all the latest stuff.)

on the other hand this shit has convinced me of the existence of the human soul, something about seeing a person in a commercial who you suspect may not be a real person sits really badly with me on a spiritual level.

our brains were already pretty good at identifying things that are slightly Off, like there are plenty of people who can tell the difference between digital and analog audio, or film. and in that case the digital recording is an approximation at reproducing, whereas AI is approximation at creating, which is gonna be way more Off.

I worked for a company that would send out this monthly publication to like 10,000 people, plus all the employees. it was pretty dry stuff, mostly about the industry and new products and what not, and it was slammed with stock images cover to cover. I thought that was weird, because this was one of those companies that tried hard to differentiate itself, why send out something that's so bland and corporate looking? so I asked the guy in charge of it once, why not use photos of employees in here? not for everything obviously but like, there was this "ask our IT guy" column (which was funny because we had about 100) and the picture was of a very handsome man in a suit who clearly didn't work there. why not use a real one? why pay for a stock image? the answer I got back was basically what in orbit said, we'd need a studio and a lighting guy and a photographer, it's expensive to do that and this is an area corporate doesn't like spending money on.

I don't know how people at large feel about this but I hate that shit. I hate when a two-man heating and cooling operation advertises itself with an image of what looks like one of the Hemsworths with a hard hat on. I feel like I'd be more likely to use them based on an ad if it was just a picture of one of the guys. I don't care what he looks like. But at least these are *people*, what AI is doing is worse, it's not even "here's our fantasy handyman, hey who knows maybe we actually do have someone who looks like this", but rather "this person doesn't even exist at all", imo it just adds to this rapidly encroching feeling that we're all getting that everything out there now is a scam.

have they tried emailing the staff? i reckon any org with 20 or so people (men) in it will have one with a decent SLR and some cracked editing software who fancies himself as the new David Bailey

(is David Bailey still a thing to fancy yourself as?)

phone cameras even, everybody has one

i saw some taqueria or something take a photo of a poster they made. it was cute and attention-grabbing.

there was a dj a while back, man power - i'm sure he's still doing his thing but i've stopped following all that - he rode to fame and success by posting a lot of photos of hand-drawn messages and stuff talking about diy aesthetics and keeping it real in the dance music industry. and also being a great dj and producer ofc. but that always struck me as a good way to do social media - draw something and take a picture of it. no idea how the algorithm handles that now though.

I have a genuine fear of these companies acquiring too much power and feel like their founders/ceos are thoroughly misguided people (or, in a couple of cases, deeply evil), but also it's useful for cleaning up photos of record covers or other items I'm listing for sale (not altering, just improving) and also making new wallpaper for my phone or my desktop pc. Not sure what else people do with it all day every day, but it's probably not good.

No matter how well I learn to get around Illustrator, I could never land on this. I just don't have the ability. But this took about ten seconds to make with a prompt.

https://i.imgur.com/1fPg6Gv.jpeg

Map, are you talking about Elijah?

sorry, JF, but that image doesn't seem attractive to me at all. but I am not an artist and I couldn't have done much better without a few months to work on it.

https://f000.backblazeb2.com/file/jpgshared/I58y_JEr

I fixed it with pre-LLM slop methods:
https://www.photopea.com/g/I58y_JEr

Pirating Photoshop to make crap to send to your friends was such a major part of Web 1.0-1.9, it’s a bit sad that adolescents will never experience that or the SomethingAwful contests.

All that also gave quite a few people useful skills.

Free photoshop alternatives are available and not at all bad - I use Krita - but they don't have the cachet.

Automatic thread bump. This poll is closing tomorrow.

No matter how well I learn to get around Illustrator, I could never land on this. I just don't have the ability. But this took about ten seconds to make with a prompt.

I understand all the arguments, but I'm neither distributing it nor selling it. It's on my computer and on my phone for me and me only to look at. But like I said, that wasn't my primary reason to use an AI anyway. It was to clean up photos. That's just a side quest whenever I want something new to look at.

Don't even know why I shared an example, but as you can see it's pretty much just a watercolor with no particular style.

I disagree that the sampling complaint should apply to this kind of picture, which is pretty abstract and might as well have been generated with non-LLM code (though these days the code itself might be made with LLMs). If anything, doing it the "old fashioned way" would more explicitly be copy/stamping random clip art and applying filters.

It is super weird though that the color scheme just happened to match elliot's hoodie and jeans.

I guess I am pretty bitter about the "AI" "industry" torching resources and ramping up fossil fuel use because people want to write fancier e-mail without effort, cheat on essays and see their family as Studio Ghibli art. Books that I took years to write were stolen for "training sets" and that's apparently tough shit, there is no higher morality than grifters wanting to get rich from government-endorsed Ponzi schemes.
My bottom line is that all use of these services is unethical. If you run LLMs on your own hardware for a useful outcome, that's cool (and I kind of do that myself for image analysis).

Map, are you talking about Elijah?

― Jordan s/t (Jordan), Sunday, August 30, 2026 7:55 PM (yesterday) bookmarkflaglink

late reply but Man Power

also maybe i'm anti-design in a sense, but give me a photo with some text slapped over it every time - no matter how crappy the photo is or whatever at least it documents a moment in space and time, so much more interesting to me than someone's "pretty" computer program doodles. i recognize that this is a niche opinion and it's not meant to antagonize johnny fever in any way, whose posts i always enjoy reading.

I still would much rather it didn’t exist as a mass market ‘product’ and think the educational, political, social and economic negatives outweigh any possible positives but I have come round on Claude being legitimately useful for a bunch of manual workplace tasks. Whether it offers good value for money (even at today’s loss-leading prices) is another question, though.

yeah that's the rub, it all feels as scammy and overhyped and terrible as cryptocurrency but it is also legitimately useful in some ways. to use JF's example, say you took a great photo of yourself but you notice your zipper's undone, AI can pretty quickly fix that and I don't think anyone's gonna give you flack for it. but it's just as easy for a realtor to use it to say make interior photos of a house they're selling look slightly bigger which I think everyone would agree is fraud.

what if your audience wants to see the zipper undone

sounds like something more up Grok's alley

Back when I got my Google/Android phone many years ago, pre-AI boom, it had a feature that could erase unwanted stuff from the background of a photo. It was a neat party trick for a minute, even if I never really used it as intended. But for sure, it's hard to argue why something like that *shouldn't* be used to remove a blurry dog taking a shit in the background of an otherwise good picture. (See also - or don't - shitting dog Trump in that World Cup team photo.)

Still using it because work makes me. However, I've been growing more accepting of it (begrudgingly) as I've found ways to use it to make some tasks easier/quicker. I distrust people who are all-in on it though - they feel culty.

It can make things easier the same way calculators make math easier. But I was shocked, just in the last week or so, at how often the (mandatory) AI google results gave me a wrong or inaccurate answer, wrong enough that it would have caused problems had I not recognized its inaccuracy. It's a problem built into the programming, the tacky aim-to-please obsequiousness of its responses. It would rather give a wrong or inaccurate answer than no answer. It would be like a calculator that guessed when it didn't know the answer.

Hate it for all the usual reasons to hate it which are correct. Don’t need it at work and every time someone volunteers that they use it for whatever in their non-work life I instinctively think “what the fuck is wrong with you”

(& also what would you have done 5 years ago and why can’t you just do that instead of using the garbage that’s burning pretty much everything down)

I just had a friend, a really smart friend, defend social media because he is a novice beekeeper and told me he could not have learned about beekeeping without the help and mentorship of people in google and facebook beekeeping groups and I was, like, people have been beekeeping for millennia! You'd have figured it out.

Back when I got my Google/Android phone many years ago, pre-AI boom, it had a feature that could erase unwanted stuff from the background of a photo. It was a neat party trick for a minute, even if I never really used it as intended. But for sure, it's hard to argue why something like that *shouldn't* be used to remove a blurry dog taking a shit in the background of an otherwise good picture. (See also - or don't - shitting dog Trump in that World Cup team photo.)

lol I'm the biggest scold for this stuff. Saw an ad once where it was ppl editing out random passerby so the picture could just be of their friends hanging out, I can't imagine wanting to manicure your experiences like that, I love to see stuff happening in the background! Even if it is a dog shitting, that is part of life's rich tapestry, as Werner Herzog says, YOU MUST NOT AVERT YOUR EYES.

when it comes to photography, a dog shitting in the background is always welcome imo

Unless the bubble bursts or something else changes, I'm going to end up in a place professionally where I can't refuse to use it and keep my job. This is a good reason to look for a new job imo.

Don't worry, by then AI will quit your job for you.

Speaking of jobs... LinkedIn has announced that starting in September they're going to be sunsetting their traditional job search tools and going over to 100% AI-assisted job searching. Hooray!

It can make things easier the same way calculators make math easier. But I was shocked, just in the last week or so, at how often the (mandatory) AI google results gave me a wrong or inaccurate answer, wrong enough that it would have caused problems had I not recognized its inaccuracy. It's a problem built into the programming, the tacky aim-to-please obsequiousness of its responses. It would rather give a wrong or inaccurate answer than no answer. It would be like a calculator that guessed when it didn't know the answer.

― Josh in Chicago, Monday, August 31, 2026 10:27 AM (one hour ago) bookmarkflaglink

its really shocking to me that Google is fine with this. I think back to my comp sci classes in college and how there were these very fundamental principles like "a specific input should generate the same output every time" and "if your results are inaccurate even 0.1% of the time then it is bad software". Google has got to have some of the best engineers on the planet, of course they must know this, they'd also practically gotten their search down to a science where the results at the top of the page were almost always what you were looking for, the fact that they threw that out for something that often gets easily verifiable things completely wrong is really something

If you use google you need to put “-ai” in the search terms to skip the worthless summary, it’s second nature to me now (tho I don’t use it a lot anymore, I use ecosia (it’s rubbish))

https://www.wheresyoured.at/the-men-who-killed-google/

They knew.

If you use google you need to put “-ai” in the search terms to skip the worthless summary

The other trick is to type in "Charles Darwin" or whatever, then go to "More > Web" and you're back in the olden days of Google

If you use google you need to put “-ai” in the search terms to skip the worthless summary,

How does it work? Eagles-ai?

DON: Well, yeah.

(i think you need to add a space before the -ai though)

https://www.youtube.com/watch?v=v2Kco_gLBkA

Unless the bubble bursts or something else changes, I'm going to end up in a place professionally where I can't refuse to use it and keep my job.

I think our current moment is comparable in this way to 1995-2005, where suddenly a lot of people who didn't need to be familiar with computers or software found themselves needed to have that as part of their resume. And similar to that, in 1998 it would make sense to a bookkeeper or architect or editor or teacher who preferred to work with pencil and paper. I knew a number of these folks then who refused to get an email account, or thought it ridiculous to "talk to people on the computer" like we're doing here, and, unless they were close to retirement, it behooved them to at least get familiar with the new tools.

i don't know, though, I don't think AI has the same learning curve/paradigm shift thing as the PC, it's not like we call couldn't use these AI "tools" if we wanted to, they're not hard to learn, in fact isn't the whole point to eliminate uhhhh having to learn stuff?

Well I could honestly put on my resume that I know how to use a chatbot, who tf doesn’t

xp exactly

The moment may be comparable in some ways, but the technology in this iteration seems to be qualitatively different in how it will affect society when we replace the older tools with these newer AI tools. I'd love to replace that "when" with an "if", but unless AI is treated with the same level of visceral dread as nuclear weapons, I can't see its global proliferation being even marginally restrained.

I had to sit through hours of THEE most idiotic "AI prompt engineering" recorded trainings that would only be helpful to someone who never learned to use spoken language. It was on the level of, in regards to prompt creation, "What if you just say, make me a picture of a house? What if that house isn't what you had in mind? Here is how you tell it what you want to see, you say, 'Show me a picture of a house with windows and doors, at least two stories tall, with a pointed roof, surrounded by a small amount of grass-covered lawn.' When you learn how to make requests clearly, with context and specifics, you will be able to generate good outcomes from your ChatGPT assistant."

Barf x100

the difference is most of the tech changes from 1995-2005 were a lot more obviously useful. cell phones, text messaging, TiVo, the iPod, it was all the subject of a lot of mockery at the time with people taking principled personal stands to avoid it for whatever dumb reason. but at the same time it was very clear why this stuff was good and what it would allowed you to do that you couldn't do before. AI is having a much harder time making a case for itself though and the public certainly seems unconvinced which makes it feel a lot different. the iPod wasn't forced on us, cars and computers didn't lose their CD drives overnight while record labels started making it way harder to get physical products, that happened because consumers actually wanted it

My colleage who uses AI for image generation legit has like a 7-8 page prompt environment of things that she has established must be true in order to get an attractive, photographic quality, non-branded, communicative image of a can of tomatoes, or a package of infant diapers, or a quart of curry powder. Ultimately it works but...it's just so stupid.

We have to have copyright free images, we don't have access to manufacturer data and we aren't allowed to pull images off of mfers' websites, before you ask why we're doing this charade.

I have a friend who's a math genius who mostly works at a national lab but also teaches at the state university, and he teaches a whole class on "prompt engineering" — teaching people how to think like an LLM so they can more effectively tell it what to do. It sort of blew my mind that this is like a 300-level college course.

genAI images are theoretically not copyright free either though, it's basically laundering copyright. dunno what is the current status of legal challenges to it on this grounds

xp Do you think a class like that is effective? I was so confused...have the people who are conceptualizing this skill set literally never communicated with other humans? This is what happens when you defund liberal arts tbh.

AI is having a much harder time making a case for itself

The literal million dollar question is not "would you use it" but "would you pay for it." That's a tougher needle for these tech nerds to thread. I would (probably) pay for a service like gmail or maps if they were not free. I would not pay for Gemini or whatever. Which probably means they are going to eventually incorporate maps and mail into gemini and make you pay for it, lol.

Certainly not many people would use it if the AI companies charged enough money to break even. Never mind charge enough to make a profit or start to recoup the hundreds of millions spent on LLMs so far.

yea kind of a big problem for them, most of the stuff it does people wouldn't pay for, and the stuff people would pay for are things that are probably highly unethical or illegal, feel like studies are consistently showing this stuff doesn't actually make people more efficient but it does make them dumber, it's just a fundamentally bad product

I’ve heard a lot of well-placed people say that there is basically no consumer model of AI that makes any kind of financial sense. It’s a B2B service. idk if I get $3k of value from it (current rate), I know I don’t get $10k of value from it if that’s what they’re going to want to charge a couple of years out.

have the people who are conceptualizing this skill set literally never communicated with other humans?

I gather that the problem is that LLMs don't "think" like humans, so humans have to learn to communicate in ways LLMs understand. It reminds me a bit of the story about Bob Dylan and the drum machine — "I have to keep up with it? It won't keep up with me? Then what good is it?"

I learned about the idea of the "reverse centaur" from Cory Doctorow's recent spate of interviews for his new book and on the whole I reject having to learn to "think like" something that doesn't think and that stole everything it knows from people.

My 11yo kid and I were just at the Dr’s office. Inbetween seeing the tech and the nurse, kid asks where Lyme disease got its name. I dunno, so when the nurse comes in we ask her. “I don’t know, that’s a good question! I’ll tell you how to fund out… [here I’m thinking she’s going to recommend a cool kids educational website or something] …AI! Use AI, it’s the first result on Google.”

Kid and I exchange a glance and look it up on a medical website. It’s named after the town where it was discovered. And the kid has covid.

my son is also 11 and it's funny how he and his friends have all independently come to understand how untrustworthy this stuff is. because kids in general are really bad at writing search queries and they tend to ask a lot of loaded and opinionated questions so they've all had the experience of AI getting things dead wrong in real obvious ways. and it's led some of to them adopting "AI" as slang for anything that's fake or obviously wrong. not really a good sign for the future of this technology and I do wonder a lot about how they're going to view technology in general as they grow up.

kids are excellent and apt bullshit detectors! Think of how often people lie to children, and how adapt they are at identifying half truths and patronizing and manipulation. And… That’s like AI’s whole jam. Conversely, the biggest AI boosters I know are insecure bullshitters who desire adulation and external validation.

*adept

Conversely, the biggest AI boosters I know are insecure bullshitters who desire adulation and external validation.

This, so much. It's making people worse people. Like online gambling, maybe I wouldn't have, you know, named my firstborn child after some people but seeing them get into this is like...I'm never going to be able to respect them now. They've outed themselves.

You know how you'll be in a conversation about, I dunno, elephants, and somebody will say something weird like 'man, have you ever seen an elephant go pee? just gallons and gallons of pee, enough pee to wash your hair in' and you try to shrug it it off because it's an uncomfortable thing to say, and you forget it the second or third time that person brings up elephants or pee or shampoo, and then you start to think, oh dang, we got some kind of fetish situation. well, that's how it is for me with AI boosterism. i was in a meeting today and a person brought up 'innovation' and 'embracing new tools' and 'labor-saving with technology' i thought next think you know we'll be talking about babar taking a shower :(

itsacult.jpg

Automatic thread bump. This poll's results are now in.

Worst part about this report? No screenshot of the CITH AI image.

https://www.youtube.com/watch?v=_4T__tIgKCU

Different state, same trick. Grown-ups need to take a long, deep breath.

https://i.ibb.co/KzxDkhKS/Screen-Shot-2026-08-31-at-9-22-10-PM.png

March 2 is going to be wild next year.

This, so much. It's making people worse people. Like online gambling, maybe I wouldn't have, you know, named my firstborn child after some people but seeing them get into this is like...I'm never going to be able to respect them now. They've outed themselves.

― Ima Gardener (in orbit), Monday, August 31, 2026 5:09 PM (four hours ago) bookmarkflaglink

online gambling is actually a pretty good microcosm of this whole thing. poker specifically. where it's at now is nothing like what it used to be back in the days when you would see Party Poker ads nonstop on TV. people assumed it was scammy but by and large it wasn't. the bonuses they gave out were legit and very easy to clear. and each site had a bunch of skins, every month you could bounce money from one site to another and collect a free $100 while only generating $10 in rake. essentially, you could scam *them*. also the action was very good, everyone sucked and some really awful players made a ton. now there are solvers and botfarms and colluding rings, they let people play as many tables as they want so all the best players are at every table, all the bonuses and promos are incredibly finetuned to make anything good out of reach. these sites are really good at making people believe they can make money but very few people actually do.

To answer the poll question: I dislike or hate gen AI but I use it because work makes me. We've been using it exclusively for a project to migrate a large codebase from python 2 to 3, to cut a long story short basically we have no option, we simply could not do it manually. But the whole thing makes me feel so queasy. The hard rule of "human review all ai generated code" is simply not possible because of the quantity, so we are entirely trusting it to mark its own homework.

Often it will ask a question - always in incomprehensible claude speak: does the H8 rule permit a scratch-copy egg_info metadata build of the post-defang aolpy setup.py? Only blocks S5-W71's gate shape, late in the stage. If no, the gate degrades to AST + marker evaluation.. So I ask it to explain in plain english, which it does but it's still talking about systems and processes I hardly understand so I often have to trust its "recommended" decision. And it's hard to trust when the output of each step usually contains an admission that one of its previous judgements was wrong.

I feel out of control, trusting something I do not understand, and I simply do not enjoy the cognitive dissonance of having a conversation with an entity that I know has no actual understanding.

That's a really weird use case since python comes with a built-in migration library. Has it become practice to skip that and rawdog it with LLMs?

No, we (or claude) use futurize, but the problem is all the behaviour changes with strings, unicode, bytes, etc. That requires real code changes and we don't have the resources to go through all the 750,000 odd lines and see if they need fixing, or the unit test coverage to give us confidence in the changes.

I feel like there's a real boat being missed by not using LLMs to refactor projects in such a way that they are more humanly manageable and verifiably understood instead of doing this black box spray and pray approach. I'll see vague posts about resurgence of verification languages etc... but I mean more like "make this project not such a big messy pile please, thanks"

Like the number of lines of code should go down ideally...

we have scoring models written in Python that we then now port to Java using Claude. thus far none of them have managed to work correctly out of the gate and it usually takes several days of us banging our heads on the wall to figure out why but that's still better than doing it manually I guess

I still hate it, and it still came up with 7 warnings when I asked it to review its own fixes, and it still said "oh I made a mistake last time" every time I asked it a question, but Claude just fixed(*) a bug with word document generation involving wordML that I don't know how we would have dealt with otherwise.

(*) hopefully not being premature, it's not deployed yet.

team lead made the mistake of mentioning that i might like to use Kiro, which i think is the only one that's officially sanctioned here, and got 10 minutes of me telling him why ai is bad. he countered with the fact his wife uses it to plan her garden.

Sci-fi guy Jeff VanderMeer offers this clarification on his social media: I want to apologize for calling generative AI users sloppers... It was wrong of me and I feel bad for the misnomer. I meant to call them "fucko demon sloppers." So sorry.

lol

the silver lining for gen ai for me is that it's interesting to learn a new kind of fake. it makes me appreciate the richness of a simple photograph more than ever. computers just can't touch the magic of our visual communication.

Liz Phair played the Greek last night, headlining after Sleater-Kinney. apparently throughout her performance, projected behind her was generic AI slop visuals. how hard is it to just use a photo slideshow or some random video montage?

Trollin' since '03

okay now its personal

https://bsky.app/profile/orbgazer.bsky.social/post/3mv4gkuj3gc2w

FYI, "This author has chosen to make their posts visible only to people who are signed in."

I'm signed in on my phone but I don't use Bluesky on my computer and I like it that way.

https://skyview.social/?url=https%3A%2F%2Fbsky.app%2Fprofile%2Forbgazer.bsky.social%2Fpost%2F3mv4gkuj3gc2w&viewtype=tree

good luck world

something is wrong

https://i.ibb.co/FkGgW4DF/Screen-Shot-2026-09-09-at-7-46-20-PM.png

HAHAHAHAHAHAHA

Wanna know how to make a party with just one Geddy Lee even better?

do you prefer john petrucci's work with king crimson or his work with bowie

thats not exactly the greatest photo of Stsven Wilson

I think you mean King Crimsen

xpost

Thumbnails are amazing esp. Steve Howe

I’m Rebert Fripp

there's a new taco truck in town that's just smothered in AI graphics and it's almost blindingly ugly. the guy on the truck looks like Mario and the girl is a yassified Frida Kahlo. it's all cartoons but its a cross between 3 different cartoon styles, just the blandest most personality free shit and you're starting to see it everywhere now, this sucks

Yeah, a restaurant near where I work just plastered their windows with AI depictions of all the food on their menu and it's just... not only is it completely unnecessary visual slop, all the food looks nauseating in AI! Like, wtf.

lol Times Literary Supplement

Commentary|Essay

AI and the crisis of the author
By Anchit Sathi

September 4, 2026
Read this issue

Editor’s note (September 21, 2026): We have removed this essay. AI was used in its preparation. The TLS editorial guidelines prohibit the use of AI in the writing and editing process. The author states that he was unaware of the policy.

https://www.the-tls.com/regular-features/commentary/ai-and-the-crisis-of-the-author-essay-anchit-sathi

lol

AIception!

I've been getting a bunch of ads for this app that can scan the label of your record and tell you what it's worth. Probably targeted towards me because I spend so much time on Discogs.com (worth mentioning by the way that the Discogs app has a bar code scanner that already does this). They're all AI-generated and full of uncanny valley artifacts (at one point the guy has 3 hands) but what was weird to me was in this ad the record the guy finds while crate digging is Spirit Phone by Lemon Demon. Not exactly an obscure release but I'm curious why of all records they chose that one - usually in these sorts of ads it's Beatles/Floyd/Beach Boys/Zeppelin etc., because those are ones you actually find in thrift stores, records that do have hundreds of different pressings, some of which really are valuable. You'd never find Spirit Phone in a Goodwill or garage sale. Not only that but it generated an AI cover for the record as well, in this case a phone with bats flying out of it. Anyway I'm starting to suspect these AI advertisements are starting to get hyper targeted in this way. I do happen to own 2 copies of that record and I have them both in my Discogs collection, I wonder if that's why it picked that one? I dunno, very very strange, getting hard to use social media at all without getting completely freaked out these days

Greta Garbo is BACK and hawking ball bearings

https://www.youtube.com/watch?v=JbttlHN0A4U