I've been very busy lately with the development of a generic chiptune player for our beloved Atari ST. And when I say develop, I don't mean that I developed it like I did in the old days. Actually, all the code was written by Opus 5 under my 'guidance'.
Now, does that make me feel proud? How does this actually make me feel? But more importantly, how does that make you feel? I know that there are sceners out there who don't want to touch AI, or anything that is 'tainted' by it. Or at least, they draw the line right at retro computing. I hope you still want to speak to me.
What's the point in using AI, some say: the whole fun of retro computing is living it like the good old days, you know, by writing all the code yourself.
Yes, writing code is great fun - I get that - but I also happen to like chiptunes. But I suck at writing chiptunes.
Don't get me wrong though, code can be a beautiful thing in itself. But nobody cares to see my beautifully crafted code (or wants to). They also cannot hear my code, or feel it, or anything like that. I've also had managers in the past say hurtful things about my work. Like, who cares about code? If it works, ship it! (for some definition of 'it works')
Does code really matter? Did it ever? Does it matter now?
I do strongly believe code does matter. Actually, I deeply care about it - as a means to an end.
Here is a bit of an anecdote: some of you sceners may know me as Rapido. I am the author of the so-called '3D delta polygon' routines. I had this idea when I was waiting for the last train that never came, so I had a lot of thinking time. But as they say, ideas are cheap, execution is what matters. And oh boy, the development of delta polygons was a real time sink.
Looking back, I've done maybe six productions in my entire demo 'career', and that's not a lot. I did have a lot of ideas (really!) that never came to fruition, because... wait... this idea is a three-month project - let's just drop it.
But now, with AI, I can finally execute all those ideas cheaply! And that is great, right?
Right? (AI psychosis kicking in.)
OK, which ideas should I spend my precious tokens on? And does it actually matter? Because nobody cares, nobody cares about code.
And then it hit me. What if I put my skills (and +200 hours of 'prompting') to good use? And by good use I mean: create tools for artists, so they can create their beautiful art.
The bottom line:
Are you a chiptune artist? Then you may like what I've built. Coming soon....
ย
Comments
you need to drive the horse :)
again treenious, hilarious - and i mean that. please share your prompt!
I think it's a great move to openly share your ideas and approach. Not everyone has the same story and perspectives differ, so your contribution may be helpful to learn living well with all of them (especially for guys like me stuck in the "let's do it like in the good old days" camp).
Please consider contributing a serious article about your views instead and send it to info
I fully agree with Terence, you got to drive the horse and it relieves you from a lot of the heavy lifting.
Claude is wonderful for gathering information, code review, writing boiler plate code, writing unit tests, testing out ideas, execute and profile code to find corner cases and just being a sounding board for ideas.
But it is still me designing and crafting the program according to the vision I have. It's still me structuring the code and dividing it into parts that I can feed Claude and review individually. I still make sure to know and understand every part of the program.
I feel like I have moved up from being a lonely coder to being the designer/architect/senior developer who is commanding an army of developers at my disposal and it has brought the fun back, since I don't have to do all the grunt work. I still hand code portions of what I do, but Claude is there for the rest.
AI assisted coding is not the same as AI Slop and I think people will start to understand the difference. I'm convinced AI will be a great thing for the retro scene. Most of us still have the passion but we are not teenagers anymore and don't have the time or energy to everything by hand. With AI more of us can produce great new products for our beloved old machines and share across the community.
Discuss, with critical reference to the definitions of "hobby", "interests" and "enjoyment".
interests: atari st, programming languages, beautiful code, helping artists, tool making.
enjoyment: fellow sceners/mates. nice chiptune sounds. not fighting assemblers or c tool chains (the grunt work).
Anyway, here is my not quite so elegant way of explaining it:
I love to do my hobby when I have the time and energy for it, which unfortunately is not very often nowadays. Everyday life stress with family and obligations as well as age takes its toll. Most hobbies also includes grunt work. If you love to grow a garden you have to spend a lot of time uprooting weed and such or cutting the lawn. Does it become less enjoyable to grow a garden if you can offload some of that work to someone else or machinery? Maybe for some, but others might enjoy being able to create and maintain the garden they envision and focus on caring for the flowers.
My hobby is crafting interesting and well thought out software. I take enjoyment in seeing the vision become real and being able to offer something to a community I enjoy being a part of, hoping that it brings them happiness and enable them to bring their visions to life (I primarily make libraries and tools for other developers). Although I enjoy the process as a whole, it brings a lot of tedious work, which can be too much of a burden given my current situation. I can now focus on the interesting bits.
With some help of AI I can do more with the little time I have and I can even make it better, which is very satisfying. I have my hobby back, which I earlier felt I had lost to lack of time.
If you're going to post pro-AI stuff, given how divisive the topic is, where's the engagement with the actual concerns that people are raising in a demoscene context? The AIscne.org pastebin above does a nice job of raising some important questions which I'd be interested in your take on, if we're discussing this seriously.
i do not see any actual and explicit concerns being raised, at least not in this thread, although of course the concerns have been raised elsewhere.
that said, my post was also all over the place - so i take a hit too.
you mention 'outsourcing', having no energy, don't have time etc. well not in my case (which is anecdotal of course).
currently, i can tell you first hand that i'm spending huge amounts of effort in my retro computing projects, even more so than ever. is that because of ai?
i don't think entirely so. i was spending a lot of time last year, but was spending that time on... boring stuff like single stepping the debugger to catch some weird stack overflow issue (rites of passage).
for example, i tried to reverse engineering chipmon ii so that we could put it back into scavengers hand. this would have been a big project, fun yes, but a bit tedious just for the goal to have scavenger do a chiptune again.
now, am i defending ai? or rather my usage of ai? i guess i am, a bit.
to be honest i also don't know exactly how to deal with it - my post shows that i'm a bit in a shambles about it.
but i am having great fun - that is my hobby!
I'm definitely in the "significant dislike of AI in the demoscene" camp - even in relation to tooling. But the first thing to say is, this doesn't mean I dislike people who use AI :). What it probably does mean is that I value different aspects of both the process and outcome to some of the commentators in this thread.
I think AI is polarising and those polarities are interesting. Even though that last pastebin above is LLM generated, it actually nails some of the dilemmas.
For instance, some people are interested in demos because of what they show is possible on a machine. I guess that position would not rule out AI use.
Others (including me) are interested in what humans can show they can do on a machine. I realise this is a subtle difference, but it's really distinct. As someone who feels this way about demos, AI use of any sort turns me off. I want to be able to look at a demo and "know" that a person has made all of it to be interested and engaged with it, especially in a compo context. It's that ingenuity that fascinates me. An LLM generating code from prompts loses me on a human level. This links to the next point.
Some people are very focused on "product". The code is a means to an end. Obviously for those people, there's no immediate reason not to use AI (unless they have other objections to it).
I really see demomaking as about craft and process. The slog, the grunt work, the "endurance" is part of the craft - and back to my first point, part of the human interest and sense of wonder for me. I can't resolve this position with AI use, which has automated and arguably, I would say, "stolen" the craft.
So people can do whatever and I'm not saying one position is superior to the other. But where there's a challenge is that whilst my (human, process driven) approach is unlikely to "offend" an AI advocate, my stance on demomaking means that AI use sits really badly with me. I think this is a really common miscommunication in these kinds of discussions, as people who use AI obviously don't experience a negative emotional/values-driven response to either its use or its lack of use, whereas people who are opposed to it do.
The final point I'd like to raise, which extends this, is that I really wish that AI users in the demoscene would stop writing things seemingly aimed at non-AI users trying to justify their actions. Because it's not going to convince anyone who has concerns about AI, as they're likely to be coming from a really different position in the first place. Even worse, I think some of these articles (which like I say don't ever really engage with the concerns in my experience) inadvertently are pushing people away from demos. Sadly, I feel this myself, and I know others do too :).
now that is a reply i like!
all the points you raise i really resonate with.
yeah, ultimately it is about humans - why we want do things - the hard bits - the breakthroughs - the stories - going on a long journey with your mates - the process itself.
fyi my article was not about convincing anyone to use ai but more about my struggles *and* what i (think i) get from it.... and to get this reply from you (i mean that).
i do experience the negative side of ai. there are obvious big negative sides (the power consumption, the theft, the deskilling). and if you really think it through, are there any noteworthy upsides?
because today it is really the wild-west and an arms race. what i would like is for ai to be regulated. for the theft to be compensated. for tokens to be taxed. for agents to pay income tax. it would be nice if we could turn back to the old days.
now let me conclude with your point about ingenuity. i think it all boils down to incentives. interesting novelty still comes from humans i think, at least novelty that matters to people.
sure the navier stokes problem has been solved. but who raised the problem in the first place? ai doesn't care about problems, unless we raise them as interesting to us.
and that is the future for me. to think of cool new stuff. for me the a demo is the next cool new idea, not so much the execution - but that's my personal take.
cheers!
I can now better see and understand your position and why it differs so much from mine. Although I have a different point of view, I respect your position and your points do resonate with me as well. I will spend the rest of this comment to share how I see all of this. Since this is a sensitive subject I apologize beforehand if some of it rubs someone the wrong way, that is not my intention.
My point of view is that I'm an alt/retro-computer enthusiast more than a demo enthusiast. I enjoy seeing the Atari scene (not restricted to Atari DEMO scene) continue to evolve with new hardware and software, be it productivity software, utilities, games or demos.
And that makes all the difference. My point of view makes me very enthusiastic about what AI can bring to the table since it allows me and others to develop more software for our small eco system and help it thrive. That's why I encourage fellow Atarians to give AI tools a try and see what it can do for them. To me it is just another tool to raise ease of development and therefore also productivity and the fun of development. So, as an alt/retro scene enthusiast I say: use it where it makes sense to you.
Demos are not my main interest and when I do think of demos I think of them more as expressions of ideas and art that explores and pushes the boundaries of what is possible on the machine, than the result of hard effort (the two ways of looking at demos which you so clearly identified).
Therefore I also see AI in the demo scene as a catalyst to allowing more people to express themselves through the art of demo making. Even with the help of AI you still need understanding of the machine, fresh ideas and great skills to make something that stands out.
The way I see it, AI will change the game of demo competitions in a way similar to introducing bikes to the sport of marathon running. Suddenly everyone gets further faster, but the most well trained athletes will still outperform the rest if they choose to embrace the change.
I fully understand that this is an unwelcome change to many in the scene, I get it and I don't want to gloss it over. This change is causing pain and loss. I don't say that this is the way it should be, but I don't see how it can become anything else, the genie is out of the bottle, and I want to focus on the positive aspects of the change. New tools such as cross-compilers and emulators have already changed the game, making it easier to make wonderful demos than in the old days. Although I haven't written any demos myself since the 90's I have enjoyed seeing what has been produced and how new tools has improved the craft. What we are now facing is a change along the same line, but a more profound one, possibly even a seismic shift.
In the end I think the demo scene will end up being somewhat altered but fine, if we can embrace the change and focus on what it brings to the table more than what it takes away. I'm not saying that will be easy but I believe it to be the best way forward. It is a natural human tendency to hold on to what has been and resist change, I guess most of us are doing it to a degree by being in this community ;)
But it would be very sad if wonderful, talented people decide to leave the community because of this shift instead of embracing the possibilities it gives them to amplify their talents and take things to new levels. That is, I guess, the reason I'm being vocal about all this, I want to share a positive message that energize people among this doom and gloom I see.
I have worked alone on my work of passion, a cross platform UI Toolkit, for 26 years now, sinking thousands of hours into it. Over the last few years I have felt more and more that I will never be able to finish it due to lack of time. The latest AI models have been a godsend to me, lately I solved a task in just one evening session that earlier would have taken me hundreds of hours. I now feel energized and back on track, hopeful that I can bring my vision to life in the not too distant future. That's why I'm so positive to AI and why I encourage others in similar situations as mine to give it a try.
Sorry for this long rant. This is just me sharing my point of view and offloading all my thoughts in written form. Hopefully some of it will be useful to others, or provoke (in a friendly way) replies that will grow my understanding.
Peace and love to you all. :)
But I wouldn't enter an AI-assisted demo to a democompo, because ST family coders have decades long traditions to craft "impossible" things from thin air. As someone with no programming skills, it amazes me still every year.
We all grow older and have various things to take care of which reduce the amount of time for our hobbies. If modern technology offers a helping hand, why not use it to keep Atari ST compatibles alive and interesting.
It still requires a lot of expertise, skill, good old manual testing and bug hunting. So every time I see someone making a YT video about something cool "they" created, I have zero expectations to see the final product.
I wonder, however, is this really *worth* to mention? To me, the discussion seems less about the fitness of AI-generated artifacts or metrics of efficiency but fundamentally about what such artifacts evoke in us humans.
I get the impression that the creators/generators of such artifacts feel the rush of accomplishment, which is relatable. To others (such as myself) the same artifacts feel dead and evoke a feeling of sadness. With the human factor diluted, all the awe about how much human ingenuity, love, and passion (tedious work) went into a production just vanishes. The story disappears. What remains *in me* is largely indifference.
Even though I'm incapable to feel it, it is good to see that those tools can spark joy when used as catalyst to revive long-hedged dreams. So I'm not here to complain, but merely to voice the saddened perspective of my own.
I fully agree with your statement. A personal experience to back it up:
I had Claude generate the bitmaps and scripts for a new theme for my work in progress GUI Toolkit. In less than an hour it had generated something that mostly worked, I would say it was 90% there.
Now I have spent another one and a half day to try to get it to 100% and I'm almost there, a few more hours work and it should all be done. That's one and a half day of prompting, waiting 10 minutes, testing the result in a repetitive cycle throughout the day.
It is not that Claude is stupid and don't understand my instructions, he really does a great job and the result now looks better than what I would have been able to pull off without him, it's just that getting the details right takes time and I do need to steer him to get a coherent and seamless end result.
In this case it would actually have been more fun to do it myself, I love playing around with graphics and designing stuff. Typing instructions to make changes and waiting is way more boring, but now I get it done in days instead of weeks and the end result is better.
There will be a lot of AI Slop, people just prompting once and releasing half-baked stuff. But half-baked stuff from sloppy developers is nothing new. The good part is that serious developers who have a unique or great vision and care about quality now can take on projects on their own that earlier needed a whole team.
I would argue that the amount of human skill, effort and passion put into these projects are the same as before, it's just that the projects become way larger.
Not saying that anyone who disagrees with me or finds it sad how things change because of AI is wrong in any way. I also see the downsides to how things are developing and hope that some things can be kept the same.
But I also feel hopeful that people who put in a lot of skill, effort and passion in their projects will make them stand out. If somebody makes a new large application such as a DAW, pixel editor or videoeditor all by themselves even with the help of AI, that will be damn impressive and the result of a huge effort. And I believe people will start to realize that.
That being said, no doubt LLMs are very useful to produce output in contexts you don't really care about, in all the bullshit part of our respective jobs. Better let it be done by an expert bullshit generator.
When talking about hobbies, it's something else. Someone who loves mountain climbing will study and organise routes, physically prepare themselves, etc. Or instead, they also can hire a helicopter and directly go to the mountain top, avoiding a lot of hassle. If you're only interested with the result it's fine, but you're not in the same category as actual mountaineers.
I kind of agree with No's perception. A product made by literally no one feels empty, soulless. It's fine if it's a banking app, because I just want to get stuff done, but about something around my hobby area, that emptiness feeling is unbearable to me. It can be amazing, beautiful, useful, whatever. It can not compare with something made by a dedicated, passionate artist.
The anthropomorphism, along with seemingly a genuine belief from some that LLMs "understand" or even "hold new knowledge", is indeed really disturbing. As humans, we're just meat and neurotransmitters; we've spent millennia evolving biological mechanisms to make sense of one another and our social worlds. Having these mechanisms be so perfectly triggered by non-social agents feels immensely psychologically dangerous.
I've already met too many people who, particularly over the past year, are shuffling off the edges of shared reality/intersubjectivity due to this. Crazy.
I don't know about others, but I refer to Claude as "he" in the same way that a sea captain refers to his ship as "she". Same is common with other vehicles such as cars, motorbikes and airplanes. I guess assigning it a bit of personality comes natural when you work very closely with it. Doesn't mean I don't understand it is just a machine.
Quote:And what make you think that I'm not dedicated and passionate about my project? Just because I use more powerful tools now to get more done in shorter time doesn't mean I'm just as passionate about my project as when I started it over 20 years ago. I think working on the same hobby project, slaving away ~10.000 hours in front of the keyboard since autumn 2000 to realize a dream and a vision without any consideration of future profit proves my passion, dedication and effort.
I'm doing it for the love of my craft, to see an idea come to life, to create an advanced piece of software as well designed and polished as I can. This is me expressing the artistic side of me as a software developer even if I don't write every single line of code myself anymore. I'm still as picky as before, obsessed with the code being coherent and beautiful and I review every line suggested by the A.I. modifying nearly everything before I accept it and still write a lot of it myself.
To have my life's work and passion so easily dismissed because I'm integrating LLMs in my process to be able to move forward faster feels like prejudice. It feels saddening and discouraging that some people don't seem to be able to see beyond that and might never be able to appreciate the amount of passion and hard work I put into this.
So, I'm encouraging everyone to open their minds a bit and start to consider that a product, even if developed relying heavily on LLM's, might be someones work of passion with a lot of dedication and hard work put into it.
On a sidenote: Zerkman, I'm a big fan of your ZeST project. Thanks for bringing that to the world and so generously giving it to all of us! :)
TL;DR: I don't care what you do; I care deeply about what I do.
For me, retro stuff, be it playing or creating games, coding demos, or sharing my musings, is all a hobby. Something I do to relax and amuse myself. If my work amuses someone else, good; I love that too.
My two main issues over the years have always been that I am lazy and that I have dyslexia. If you've ever read any of my old scroll texts or README files, you know what I mean. An LLM can proofread and fix my text; I love that my thoughts can now be made much clearer.
The laziness is the greater issue. Once I figure out how to do something, I get bored and move on to the next thing. The last 10% of development is 90% of the job. So I have full-screens, fire, water, and lens effects, and so much more on various disks; they're all forever stuck in the proof-of-concept stage and never released.
What I have released was saved by tooling. Image converters written in STOS, and later Pure Pascal for 3D models, timing tables, and much more; WAIT would never have been finished without Pure Pascal. And it's a miracle it was finished anyway. Typing the last 90% of the assembly code for that demo was no fun. A pure grind. Something I endured, not something I enjoyed.
When I picked up retro coding again, I went with C, then C++ with GCC and cross-compiling. That means the 90% I don't find fun is greatly reduced, and I can spend my time on the 10% that brings me true joy. I see LLMs as the next step here: the most tedious parts of the 90% I do not enjoy can now be outsourced to an artificial intern. A very fast, enthusiastic intern who needs quite a bit of supervision.
If I may add one thing here. Reading the comments, I see two camps:
Product/idea: I had an idea. I want to see what it becomes.
Craft/process: I want to experience making the thing myself.
I put myself somewhere in the middle. I have ideas, and I want to see what they can become. But I also want the experience of making the thing myself to be enjoyable. I would never ask an AI to complete a task that I don't honestly think I could achieve myself. I recognize that part of what we do is tedious boilerplate. For me, using LLMs removes that tedium.
I can also see others who may view LLMs as a way to do things they know they couldn't otherwise achieve. I understand and respect that viewpoint as well. It is just not for me.
In the grand scheme of things, it's about doing what I enjoy most: thinking hard and solving hard problems, the problems that are new to me, learning and exploring. Whether it's a Pascal compiler, an IDE with code completion, an emulator in place of real HW, or an LLM writing my Python scripts, these tools remove tedious tasks so I can focus on the fun.
And yes, I did have ChatGPT proofread this message.
But if it's "just a tool" why do people feel the urge to praise their use, express mixed feelings about using them, or draw limits between when to use them or not? Just use them. It's fine, really. Just, be honest about what you do.
I personally use LLMs to get advice about some coding. To produce boring, boilerplate code that is useful but with 0 interest in doing it myself. Mostly for my work, but even on my hobby projects, on very specific parts that are far from the core part and Iโฏwould have never done otherwise. I am not proud of that. It might be about 1% of my code, but concerning those parts, I'm already feeling I'm losing expertise on these concerns, so I will probably keep on using these tools mostly to get advice and not directly produce code.
OBTW, this reply was not LLM-generated, -proofread or whatever. I was written by a human, for human readers.
"In the grand scheme of things", seriously. This is getting ridiculous.
i'm really bullish on NOT using a llm for human communication. if i spot you using a llm to 'reply' i will be as disgusted like you would be also be.
don't be a meat proxy god dammit!
your remark about not being proud using a llm..
first i like your honesty - and ... yes i'm also not feeling proud using them.
but those llms do get my creative juice flowing - but the new juices are very differently from previous ones ....
ideas, ideas, ideas. lot's of them - let them all out! express!
cheers
But when speaking of camps, please consider that some people may perceive LLMs as plagiarism machines, rejecting their generated output on matters of principle regardless of their utility value. When assuming this perspective, a creation that includes LLM-generated content has a similar smell as, let's say, a PhD thesis that contains plagiarized paragraphs. Once tainted, no amount of human curation will suddenly make it genuine. For readers who are sensitive about the story how a thing came to be, this can make a difference in perception. For others just enjoying what they get, it does not matter. There is no right or wrong but merely different perspectives.
The use of AI tools for learning and exploration, like what you did for finding your way through GCC, or analyzing code for bugs, or proof-reading text, is an entirely different matter. No creative output is outsourced.
Tord, I'm afraid that some principles won't bend by more open-mindedness. And that should be totally fine, expected, and respected. Please don't feel offended and continue doing what you enjoy most. Not everyone can be pleased anyway.
https://www.youtube.com/watch?v=xWYb7tImErI