Post

afterthought 1 — "Finally, art is democratized," proclaims area programmer man making $200k per year. "No longer is it held solely in the hands of the elites: furry porn artists struggling to pay rent"

afterthought 2 — clear bridge between "you can't customize LLMs like you can tools" and the almost comical unpredictability of image generation, where getting the same character twice is a crapshoot. but i didn't want to get into image stuff too much or the post would've been twice as long

afterthought 3 — i forgot i muted "AI" and some related words on bluesky so i'm sorry if i don't see your reply lmao

One of my friends - very clever guy even though terrifyingly normal (fintech programmer, hobby seems to be buying expensive things), who I used to really respect as a good programmer has pivoted to saying "I don't understand why anyone would program without chatgpt" and like, im so confused

I agree with basically everything in this article. I tried using some of these tools out of curiosity once, and because I was trying to do something very weird I wasn’t sure how to do. It sent me down a like 10 hour rabbit hole trying increasingly more elaborate workarounds which didn’t work…

sorry for popping up under an old thread, I saw the post now, but yeah this pisses me off a lot. Like the last 3b1b video, which was a guest video by another channel was about image generation...and mathematically it was fun and interesting, as the problem of making the slop machine is

actually mathematically interesting. but like at the end this guy drops the line on how this "democratizes art...you now just need natural language" and i'm like ...you just needed your wrist and a pencil? also like art is part of expressing yourself, not just making the picture!

which is also funny because just before reading this blog i learned about T. Kinkade, an american artist (that wasnt a very good person) but made very popular cozy art, but couldn´t still manage to sell more personal and weirder works, because the public was interested in the cozy "whatever" pics

really hit the nail on the head with this one. loved the carpentry tool analogy. i am waiting for the day that these companies crash and burn, hopefully sooner rather than later to minimise as much as possible the cruft of these Whatever generators polluting the internet.

One thought that solidified while reading this: maybe spending a decade telling people they needed to learn to code if they want to make a living has lead to a cohort of programmers who are less likely to want to be writing code.

To be clear this isn't a moral judgement on those folks and it seems pretty unfair for those who have invested the time and effort only for the same industry's narrative to switch to calling them suckers for trying.

I admire the way you can articulate the vague feelings of LLM-slop-related dread I have but am not quite cerebral enough to verbalize. I really enjoy your writing. Even now, when my brain struggles to focus on longer-form text, I find your posts so gripping. I’m so glad you continue to do it. 🩷

I broadly agree with this. I definitely agree using LLMs as a substitute for actually writing code is very dumb. I also had a similar experience as you mentioned about Ren'Py when I first poked around to see what the hype was about. But I would like to give more context to this -

The thing is that since then, LLMs got (a little) better, and I got (a lot) better at understanding how to use them effectively. They don't *always* get it wrong. There have been a fair few cases for me now where prompting an LLM was the fastest and easiest way to get an answer and solve a problem.

Anyone insisting you *have* to use them is being ridiculous. But they do have actual uses. And I don't mean in terms of slopping out bad code. I mean in terms of recovering just a little bit of what we've lost to the enshittification of search engines and the erosion of forums as a place for Q&A.

And all that's not to say you're wrong in the broader point. In particular, it feels like making an LLM actually useful involves struggling against proprietors' attempts to make it Whatever instead. I'm genuinely looking forward to when I can run my own local LLM with RAG and without the Whatever.

I love your writing :) I've noticed that the Whatever all ties in together - AI pushers don't mind the weirdness of AI art or the blandness of its text, because all that matters to them is the "content". It doesn't matter to them what it is - just that it's there.

This reads like people have been sleeping through mountains of shit being thrown at them (content, news, user-hostile services...) for decades, and then suddenly woke up and blame the most recent hype. Most of Internet is crap. Most of media is crap. This does need to stop you (or me) from creating

really appreciating the extremely long reply that looks like it was written by an AI, referring to you by name like they're observing you from the moon, talking about how useful AI is, because it can do things like. uh. explain what terms mean. which is something search engines could never do before

The comparisons of LLMs (in the context of programming) to tools (in the context of various industries) that actually provide value and utility and allow people to reliably and efficiently do more is kind of mind blowing to me. Lots of desperation to justify the use of this shit technology.

"I think the core of what pisses me off is that selling this magic machine requires selling the idea that doing things is worthless. Because if doing something has some value, then it must be somehow better than pushing a button and receiving Whatever for essentially no cost."

i saw the same post you saw, not the carpentry one but the “not using chatbots is very stupid” one, and indeed it made me roll my eyes so much. just. ugh.

I just noticed that the style for the tangent block's title is a little bit off on Android Firefox and thought you may want to know since you clearly put effort in expressive styling (and I still refer to the CSS Baby series whenever I have to deal with grids once more 😁).

ah damn, looks like i was doing an incredibly stupid trick with ::before and it does not work so well when the title wraps haha. thanks i'll figure something out

This is so well written, and accurately captures much of how I feel about these Whatever technologies and the industries that peddle them. Well done. And the “spinning whirling mass of blades” analogy made me laugh aloud and wake my spouse 😆

I really like the description you've given here, the fact that it defaults to lying is just so useless. I particularly like "Whatever is something", and that whole paragraph. ime I've found it useful for really short easy-to-check code (< 2 lines), but I like autocomplete. I can get disliking that.

what scares me the most is that LLMs are the worst possible way to externalise knowledge. i don’t like admitting it, but i’m dependant on search engines to find information i would’ve had memorised otherwise if they didn’t exist. being dependent on a smarmy lying chatbot sounds terrifying

This might be the longest piece of writing where I've agreed fully with every single sentence up to the end. Thank you for articulating all of this so clearly (also i hadn't considered that link with the hopes of bitcoin also being sunk into Whatever)

Also, “What I *do* know is that a table saw quickly cuts straight lines” is such a mood, especially when talking about computers: those machines are made to deal efficiently with precise data, why use them to convert data into untraceable sludge -_-

wow that's a whole thing i forgot to say actually, that computers are wonderfully precise and we keep making them worse at it

I have a theory that it's exactly the point. I think you and I are into computers because of the preciseness of the results and what they can do. But keep in mind that there are millions of people who hate computers: because it's "cold" or "soulless"

in agreement with basically every single thing you’ve written lmao. my “fucking fine, what have you got” moment was trying to parse things in a datadog log pipeline and being given three separate completely imagined solutions.

it’s now at a point where I review a coworker’s PR on something that needs to be high performance and there’s code parsing a JSON string, taking one property from it, uselessly serialising it back to a json string, and deserialising it again using a second library.

i left a comment asking him to fix it, and he replied like he understood what i was asking him to fix, then followed up with mostly rewriting the function but not actually fixing the thing I pointed out.

I thought it was weird but shrugged it off, then I see him in the office the day after and he’s got chatgpt open on a second screen. i don’t rightly think he would’ve made that mistake before, and the despair i felt in that moment really just overwhelmed me.

One of the reasons people sneer at those not using the LLMs may be because people like me are out there actively trying to construct social mores that say it's unacceptable behavior and so they feel threatened by those not on the side of permanently switching your brain off

Feel the last section hard. Seems like a bunch of people just don't want to make things, and like, why make them at that point? Do you just want the "prestige" of having made a thing? Wild to me

They used to have this non-functional code as their main screenshot: bsky.app/profile/drac... They still have duplicate unit tests on their “Generate unit tests” page: docs.github.com/en/copilot/t... tenbluelinks.org might be useful for getting rid of the Google Search AI summary, HTH!

Matthew Somerville@dracos.co.uk

The first example on github.com/features/cop... still shows broken code (look at the return...), doesn’t include all the tests it says it does, and uses a float. I guess this is a good demonstration, but perhaps not for the reasons they want

Every word of this. Including the wasted promise of a less monopolised decentralised payment economy. Basically, we can’t just have nice things because of mass stupidity, ignorance, apathy, and greed.

The most frustrating thing about the advent of these Plagarism Machines is that my parents eat it up. My dad is (or was, he doesn't really draw anymore) an artist; a pretty decent one, too. He was there at the start of the furry fandom - even helped design the first fursuit, if you can believe it.-

He never really got anywhere, due to a maelstrom of personal problems and unwillingness to interact with the larger fandom due to ancient fandom drama he never got over, but he put his heart into what he drew. Always said his greatest wish was to be able to support us with his art. But now?-

Now, all day, every day, he sits on the sofa, giggling at AI videos of kittens that make human baby noises. He obsessed over AI audio of an emotionless, monotone voice (apparently supposed to be Batman) talking about how he'll take down [character from other thing]. I've tried to explain to him-

-why this AI is harmful to artists (like HIM for example), nevermind low-quality slop, and he just replies with "It's just eye-candy!" and the sheer mindlessness and flippancy just stuns me into speechlessness, which he interprets as winning the argument and goes back to consuming the content.

(I realized that anxiety over browsers not saving scroll position competently led to me not reading long-form content online - even when I would finish it in one sitting. I figured it was worth mentioning because I doubt I'm the only one struggling with reading in this manner.)

this is real good, really crystallized the underpinning plague of guys who don't really give a shit about anything but insist on monopolizing all modes of communication out of some vague self-promotional drive that exists beyond these fads and will survive any critique or reaction particular to them

Thanks for writing this! As a student it's been so alienating seeing my fellow students unironically regurgitating slop during group projects and it's starting to turn me off of group projects in general. It's so refreshing to see someone put my feelings into words better than I ever could 💖