Can an artist with no coding background realistically make a game using AI coding tools now?

leoma0 wrote:

Do you think AI tools are now good enough for a non-programmer to realistically make a small indie game alone?

Yes, and no. They can make a game, yes, that you could release, and players could play, no doubt about it. There are examples of games made entirely via AI. The quality on the other hand... that is a different topic. I remember a "vampire survivors" clone. It was, all things considered, a game. It was functional on a basic level, it was playable... but the "balancing" of all the random abilities with random stats made the player entirely unable to die after the first few upgrades. So in that sense, no, they cannot make an actual meaningful game themselves.

Also, ask yourself this, given you are an artist: Do you belive that it is possible for a non-artist programmer, to make a high quality game using only AI image/model generators? The answer will be pretty much be the same for your case 🙂At the very least, you will need an understanding of the systems you want to adapt, so some game-design knowledge, similar to someone using AI to produce some assets, that are then hand-edited/refined.
Coding for games is also a bit more complicated. I've found that AI is very, very good in coding websites, the whole front-end stack (where learning how to properly do that stuff is borderline unnecessary for somebody with little prior experience like myself). Games, I've personally had mixed results. I use it mostly as "just another tool" in the toolbox. It sometimes helps me out, it often doesn't. I wouldn't magine somebody using it to make a whole game, and it turn out decent. But it also depends on the scope - some simple little flappy-bird type app, could work out somewhat ok. But if your idea of "indie" is the more prominent/bigger examples, like, Stardew Valley, Hollow Knight, Minecraft... not at this point, if you ask me.

Can Amazon with its army of indians and its mandatory use of AI make a “buy” button that isn't broken? And those guys have at least a basic clue about how to program.

Modern software is already terribly bloated and slow and buggy, and modern engines already encourage building extremely wasteful games. AI cannot think, and cannot maintain anything more complex than maybe a calculator tool. It will never do the best thing, but always something that somehow barely works, uses way too many dependencies, is way slower than it should be, etc..

And while it is completely unmatched at writing code, if you look at the speed of the process, and the volume of code generated, that code is always an incoherent mess, and once it is generated, the AI is basically unable to do anything with that code again. Also, even experts who do know how to program well, have a hard time instructing AI to do anything useful in a non-trivial setting. You need to be able to read and understand code in order to work with it. And no code is correct the first time it is written, because programming is an exploratory and iterative process. But if you have 0 clue what to do to fix a problem except add “make no mistakes” to your query, then you're completely lost. AI cannot grasp and understand the code it previously wrote, and especially the more code you generate, the less the AI is able to understand what's actually going on.

The only people who can say that the AI is as good or better than them are people who suck at programming. Programs are extremely fragile, and a single implicit assumption that was made in one part of the project, if you forget to consider it while working on another part of the project, can completely break your program in not immediately obvious or apparent ways. And programs are full of these implicit assumptions everywhere, and the degree to which all this implicit knowledge propagates itself through the whole codebase also directly dictates the upper performance limit of a given codebase (alongside other factors, obviously, but if you chose to split the program into thousands of tiny isolated modules that are each individually AI-generated, your program performance & memory footprint will be abysmal).

Learn to code, find someone who knows how to, or give up. And even if you did somehow manage to make an AI-generated game, that would still be an affront to the human soul.

Thanks for the detailed reply, it was really helpful.

To be honest, a lot of artists in the game industry are already feeling anxious about AI. It has definitely reduced production time, and some people are already facing job loss because of it.

That’s partly why I started becoming interested in programming and game development. Even if AI can’t fully replace real development knowledge, I still think learning some programming is probably a good thing for artists now.

I agree that my idea is still not very realistic at the moment, especially for larger games. But maybe AI could help small teams create more polished games with fewer people in the future.

leoma0 said:
I still think learning some programming is probably a good thing for artists now.

It is definitely good to have some understanding of what the hardware has to do when you make art. The way you do your art can have a major impact on the overall game performance. For example having tons of materials with unique shaders in a game is why games nowadays need ages to compile shaders before launching and then also when loading the map, and you also get stutters whenever a new material appear. Old games didn't have nearly as many shaders as games nowadays do, in part because they couldn't afford to, but also in part because teams were smaller, so technical considerations still affected the art, and art considerations still affected the programming.

Triangle topology and density can also affect performance by a lot: it might be cheaper to have more triangles with better topology and more uniform density than fewer, but badly arranged triangles.

And shader-generated effects like smoke or similar can also vary wildly in their performance. And especially as an indie, you better target lower end hardware, rather than high-end hardware. Because if you target only high-end hardware, then you also exclusively compete with AAA games which also have the same requirements.

leoma0 said:
But maybe AI could help small teams create more polished games with fewer people in the future.

It won't. People who use AI will never polish their product, because if they cared enough, they wouldn't be using AI. All AI is doing is make people believe they can attempt projects that they cannot realistically tackle with the amount of manpower and budget/time/skills they have. Rather than living above your deserved means (just like the people who live on loans/debt as if that's the expected norm, same with companies), everyone should instead focus on putting what they actually have to proper use. Designing around hard external constraints (HW capabilities, budget, manpower, time, skill limitations) leads to more carefully thought-out designs that are also of higher quality and realistically achievable.

RmbRT wrote:

All AI is doing is make people believe they can attempt projects that they cannot realistically tackle with the amount of manpower and budget/time/skills they have.

Kinda.

For me there are a lot of parallels.

People have made banners and flyers for ages, and mass-printed them since the printing press. People used programs like Bannermania on their dot matrix printers to make banners with the continuous paper, and used clipart and Word Perfect graphics and later MS Word Art and clipart to make their fliers. They don't replace professional products, they don't replace professional graphics designers, but they're functional for the masses. Professionals can use them effectively for prototypes and quick tasks, and they can produce good results. The masses weren't going to pay for a graphics designer or pay for professionally-printed banners and fliers, this enables people.

There were (and still are) App builder tools out there, like MIT's AppInventor. Anybody can put together simple building blocks and make something that functions on their phones. A few minutes with easy-to-use tools and anybody can put together an app. They include networking functionality and Bluetooth, you can put in graphics and animations. They don't replace what professional developers can create, and professionals can use them effectively for prototypes and quick tasks, and they can produce good results. The masses weren't going to pay for professional programmers to build the applications, this enables people.

Just like "anybody can make a Happy Birthday banner" and "anybody can make a yard sale flyer" in the 1990s, and in 2010 "anybody can make an app with AppInventor", and with tools like Unreal and Unity "anybody can make a game", they are all true enough.

With AI, they are good enough tools, people with simple needs and enough time/patience can make something good enough for their needs. Professionals can use them to quickly prototype ideas and to generate ideas, but they won't replace what professionals can do with enough time and creativity and professional skills.

And as the tools get better, the professionals will advance to use them. Just like professionals no longer farm out their typing skills from dictation to an army of secretaries, but instead type their ideas nearly as fast as they can think them, so to will our game development tools continue to advance. Game engines were once enormous investments that only the biggest corporations could afford, today they're simple middleware, someone can follow tutorials and within a week produce something that would have taken a team of 50+ developers a year or more, sometimes beyond the scope of even that group. Throw in some AI-generated models and they've got photorealistic lighting effects and shadows and subsurface light scattering, the scene lit with adaptive lighting, and continuous levels of details that were only available to Pixar and others with tremendous offline computing resources, click a button to enable Nanite. Networking functionality comes free in the box, include a built-in sample and you've got a 100-player server with full position-dependent communications prioritization, click a button for CryptoPP or some other library to encrypt communications streams.

These enable amateurs and professionals alike. Yes, amateurs can now create games that took a truckload of money for paid professionals a decade ago. Professionals can quickly create demos and test levels that were AAA-quality a decade ago. Everybody benefits.

If it is enough for what an amateur wanted to create, wonderful for them.

leoma0 wrote:

Hi everyone,

I’m mainly a 3D artist, not a programmer.

Recently I’ve been trying AI coding tools like ChatGPT, Claude Code and Codex, and I’m surprised by how much they can already do.
With AI assistance, I can now build simple gameplay systems and prototype ideas much faster than before.

So I wanted to ask experienced developers here:

Do you think AI tools are now good enough for a non-programmer to realistically make a small indie game alone?

Curious to hear your experiences.

Inb4 "Omg long post = AI generated"

This will be my first and last post here. Bad-faith argumentation drives a foul first impression of this community. I am making an account to share my experiences for OP and anyone else with the same question. Take my advice at face value, or leave it.

Don't preach to me about what you think LLMs actually are - I already studied it a decade ago, long before I looked at Game Programming. It's extremely easy to iterate with AI for many obvious reasons. It's expensive and complicated to iterate with people, for equally obvious reasons.

To OP:

I have actual experience building a full stack game from the ground up with AI tooling for a game operating in stable live service. It's not the biggest project, but my 400k+ lines of code work (dependencies and all). People play it and things don't randomly break. With AI doing the grunt work, summarizing calculations, and being a sounding board - I could focus on my actual strengths in actual game design and direction. Hell - I have people hired to draw art icons. I might even get some programmers to refactor my code properly if my game makes enough money.

What you're asking is doable, but don't fall for the snake oil "it's super easy" services and apps. If you try to use AI as an effort-free shortcut, you'll be disappointed. AI requires active management. Nothing (except the rate limit of some tools) stops you from working as hard as you like in a way that actually makes progress. Nothing replaces learning basic coding structures.

I use mine aggressively - and get good value for the (cheap) plan I pay for. Ymmv but you can set yourself up for success. If it's a prototype, then the code absolutely can be a mess, because it's a prototype. I have used both AI online - and AI installed on computer. Sometimes I will ask one to sanity check the other, and yes they can catch things and think "differently". It's not hard to have multiple chats with project context for different things, referencing coding standards. Nothing but exhaustion stops you from working on your idea 24/7. A programmer will go home or space out. Your project isn't their baby.

There is no simple way to communicate what I've learned through trial and error, sorry. I'll give you an example workflow near the end.

Study the holistics of game structure so you point your AI in the right directions from the getgo. (For instance - separate controllers for different functions, rather than having everything in a few mega-files.) You need to understand the intended game flow, the systems to design, etc. The more specific you can be, the better. But when you do - AI simplifies the implementation barrier and massively speeds up your process, if you don't do too much at once. The real professionals often use AI-assisted workflows and its not something to be snooty about. Ask any experienced programmer and they'll tell you that you're doing Systems design.

AI works very efficiently given enough constraints, information, and structure/direction that you provide. Be willing to make mistakes, and investigate with your AI.. I can describe a hundred nuances of working with people AI, but you'll need to get experience all the same.
- If nothing else, you can use AI as a sounding board and an idea repository.
- Codex (on desktop not browser) can hook up to Github for version control. I use this to simplify migrating code from my personal computer to my live server. I don't know which other AI work with this. I highly recommend you look into this with your favorite.
- Make sure your AI knows your tech stack so it can help you diagnose problems or pull up commands specific for your situation - instead of forcing you to trawl vaguely helpful replies on Substacks. (They can also help you understand the tradeoffs to decide.)
- Make sure, if you don't have github, you have some version control, so you can revert mistakes easily.
- AI tends to default to the cheapest, simplest "solution" to your problem. This is absolutely fixable. Ask it to find the "real owners" of the gameplay elements or UI - so it is not making nudges to the code that are ultimately overriden. If you do not do this, you will be confused why later fixes "don't work". I've seen real coders set the same gameplay function up in 3+ different spots. Avoid.
- A new AI (without context) will default to broad theories without proper application - much like some programmers. 😉
- You will benefit immensely from building up shared context rather than starting new chats. Make an account, at the very least - or install Claude or Codex on your computer.
- AI has limited memory and will compact things over time. But you can tell AI hard rules you want it to follow. The approach, the status of the project, and so on.
- AI will not read into your intent unless it's really good and has built up experience.
- I cannot emphasize enough how much showing screenshots of my game behavior has helped the AI intuit what's going on - with codebases it is familiar with. I will mark up screenshots, make modifications in to show what I want instead. Better than pure text.
- Any decent AI will absolutely know about standard programmer practices and can find inefficient code and debug things far faster than most devs. Unlike most devs, it doesn't have to sleep.
- AI cannot invent (well) without structure. It can tell you if ideas make sense - based on what others have done, or the style of game you are trying to make - or the playstyles you want to encourage. Again - context you have to decide first.
- You can set AI conversational styles. You never get intentional rudeness from them. Ahem.
- Many of the more complex game archetypes will be a burden for you artistically, even as an artist. AI cannot fix the scaling issue. Some types are best left to studios - even small groups of people.
- AI can pull information from some documents. This varies. Codex can read more document types than ChatGPT. I use this so AI can keep an eye on my task lists and general game design direction from uploaded documents.
- AI can help you figure out obscure errors. Or - if you ask nicely - do a "deep dive" into why a bug exists. Software is complicated. There's often many factors affecting a UI element, etc. This doesn't seem important until you find bugs you'd never solve on your own as a novice programmer - that would end the game development right there.
- Speaking of which, I like getting the AI to color different UI zones different colors - this makes it exponentially faster to see "what's wrong" - what elements are misshaped or bigger/smaller than you thought - or not anchored, etc. etc.
- You will periodically have to make new chats because the chat history the AI holds can get very, very long. The "context" it builds up will bog things down. Ask the AI to send you summaries of the project state, the programming disciplines and structure you are using, the current task list, etc. I recommend making a code map you update regularly as well, so it can help you remember where to find things.

To give you an example workflow, let's say I want to add a new class to my game's "Class Selection" ui:
1. I independently build a game design document. Classes, let's say - with a simple UI for selecting and choosing. I should know what text to display. What shape and structure the class takes. Etc. You can build placeholders in.
2. I show this to the AI - and ask it to analyze our current structure. Preferably with tech stack knowledge. The language I personally use is "trace the ownership chain" - i.e. "what actually controls this element". Be specific with intent, be less specific about implementation - it might default to a more fragile method at first. (I got hung up on "tool tips" for instance.) I draw my intended modifications to the current UI and upload that in our discussion. This may be referenced many times when things don't work as expected. If you understand the basic game structure, you can help steer it in the right direction.
3. There is inevitably some discussion - asking it for help on mathing scaling factors. Or "Did you actually mean this other thing". It might place the new element at the wrong section if you are not specific enough. It is a computer.
4. Either I have an AI talk me through the process of finding my code - I show it my code structure with pasted text or screenshots - and it fixes it. Or I ask a service like Codex to write a draft for me.
5. Visually verify and test that the addition works. Does the Class page show up as expected? (Probably not.) Be specific in what does and does not work. It helps if you have insight into coding structure. If a couple quick tweaks to your new code don't work, then there is probably a hidden conflict somewhere with the surrounding UI elements (in this example). Or perhaps the UI elements are controlled by a totally separate function at the same time - and your code is being overriden. You can ask your AI to deep dive and find the "real ownership chain" in this case. Or similar language.
6. You repeat steps 4-5, and sometimes 2-3 over and over. There is no shortcut. Maybe the new image is wrongly sized (or the image atlas is being referenced incorrectly). Maybe you didn't specify the right kind of structure - or realize you made a mistake, only to go back to the drawing board. Note how an AI agent won't sigh and roll it's eyes at you for this.
7. Perhaps you managed to group a number of tasks together - in excruciating detail and specification. Maybe you asked Codex to deep dive on a bug for you all night. But when it works, it works. I can push to github - or my Codex can if I'm lazy, and pull on my live server, post testing. Make sure to tick off the completed task on your AI so it knows the feature works!

(I do recommend having a desktop AI check your code repository for dead/orphaned functions and code that can be combined or files that should be split.)

I am not familiar with how hard 3d art is, but it looks complicated to me, and I highly suggest you focus on text-based games at first, if only to learn the AI workflow. And then make sure anything you build is art-light. For 2d placeholders in my concepts, I can draw stick figures, edit images. For my 2.5d moba sample, I made hockey pucks with arms for simplistic action characters that could still express "different abilities". I imagine 3d art is a lot more work.

You need to be able to start with protoyping a simple game loop you can get right, then incrementally add on features from there.

Some archetypes are much more difficult than others. The simpler the game, the easier it is for you to structure and code. The commentary above frequently cites poor examples of AI usage and then generalizes those as being representative of the whole. I would not use AI for heavily graphics oriented games - it's far better for say - helping you manage a Ren'py visual novel or making a text-based game or just functioning as an ongoing task list. I've prototyped MOBAs and traditional Empire games, etc. with just a few characters or empire states. Nothing crazy. No complex things with dozens of characters (at first). Text based, maybe with a database, is the best case scenario.

...But it's doable. You'll run into different problems faster. But you don't have to spend years learning the nuances of a programming language's quirks through trawling demeaning substack responses to figure out why your game-ruining bug is happening.

Give it a go for free. Nothing is stopping you except hesitation.
I'm sure there were humans afraid of Fire and the Wheel.

As an aside (not to OP):

I take particular issue with snooty know-it-alls who act like anyone who can't do what they do is stupid and lazy. I enjoy chatting and helping others. I had originally intended to make an account here to ask my own questions for your "approved" programming. I've changed my mind after seeing some of the attitudes openly displayed in a space meant to be fun and encouraging. I found this attitude extremely prevalent among some of my fellow students at my Polytech, and it's a shame that it persists here.

There is no value in shooting down someone's dream and refusing to provide useful direction. You know who you are. It's disgusting and counterproductive.

FYI:
- Most people lack the funds to hire programmers to build prototypes for ideas they are unsure they should pursue.
- "Free" help has its own obvious problems. Both are challenging to schedule and maintain momentum with.
- Many things are lost in translation and correcting them can be a lengthy process. Think of software onboarding (if you've ever had a programming job).
- The most coherent systems tend to be made by fewer people. 🙂 They just need the tools to do so efficiently.
- Sometimes people are bad with syntax or learning the quirks of a software language or debugging. Perhaps they have ADHD, Dyslexia, or other challenges. No, programming doesn't just "work" for all of them. Telling them to "just learn to code" is callous and able-ist and you should be ashamed of yourselves. Do you tell everyone to move their couches alone or "just run faster"? Do you rudely ask why everyone can't dance or dress equally well? Do you tell overweight people to stop taking transit? It would be very easy to ask pointed questions based on your behaviors, but that wouldn't be helpful here.

I never get attitude problems with an AI (and they're far cheaper). If I frame my bug description well enough, the AI can absolutely succeed. I've met some insufferable programmers who blatantly ignore my design feedback as a paying customer because they think they know better. I've had programmer coworkers who were lazy with their programming architecture and hard coded placements and values - and refused to fix their mistakes despite being paid well. Suspiciously similar to the results some of you can't get an AI to correct. 😉

If I catch an AI building something wrong, it tries to fix it. No quiet quitting. No overcharging of hours. No wild bureaucratic goose chase fiddling with the place and structure of my bug reports in an attempt to keep me from doing my bloody job in catching their errors.

We have instant pots for those without the time and energy to cook "properly", and yet plenty chefs still have jobs in restaurants and in private service. We have refrigerators instead of carving ice in the winter for our cellars. We still have rickshaws despite the emergence of trains and cars, for crying out loud. The world hasn't ended.

AI is more an accessibility tool than replacement for an expert, so stop being gatekeepers. Nothing stops a user from being introduced to programming concepts gently through AI - and learning to fix things directly. Some people need the help to learn to code, never-mind using it to build things. Those videos meme-ing about teaching vibe coders to directly code would be good marketing for your profession if your attitude doesn't put them off first.

You should be encouraging people to dive into your field, and encouraging them to tackle challenges - not insinuate snidely and smugly that they're stupid or lazy for not being wired the same way you are, or for considering other approaches given their differing circumstances.

Common anxiety issues aside - it is never as simple as "if you cared you would do it." Useless Gatekeepers.

Adios.

PS I'd fix my post to adhere to the posting guidelines but your link doesn't work for me. Sorry.

I am a proud vibe coder. Now, there is a problem… the game that the AI writes is based on your guidance.

I have made 2 versions of our current game. The first was a horrible mess, and it was my fault. The second version is much better, more object-oriented, a cleaner design. And I’m a proficient coder.

taby wrote:

I am a proud vibe coder.

I'm afraid you still have a long way to go. To find out about yourself, who you really are...

Because something like a proud vibe coder can not exist. So you must be something else. ;D

leoma0 wrote:

I’m mainly a 3D artist, not a programmer.

Lean into that. It's your strength, so how are you going to leverage your strength?

"Art heavy" games have a history, going as far back as the early Full Motion Video adventure game era. Of course, that's the reason that segment of the industry imploded, back in the day. Too many art assets. Not enough adventure geek paying customers, willing to beat their head against a brick wall to make progress in the games.

Who's gonna pay for your digital art? Why will they, as compared to all the other digital art assets out there? You need to think seriously about the art marketing issue. Especially because in a lot of genres, once a player gets past the initial "gee wow it's shiny" purchase decision, the art may not matter much for the gameplay.

Heck in competitive multiplayer internet situations, people often turn down the rendering features so they can get better response times. Or if it takes a rather expensive computer to render your artistic vision correctly, people don't necessarily have those, so they turn the rendering down. If your art now looks like crap on a low end computer, well that's a lot of people who won't be buying your game "for the art".

The art question doesn't depend on AI as a question. It was already a question, long before anyone had any non-programmer tools to work with. People were debating this one back in the days of the Apple Macintosh and Hypercard. This thing called MYST came along, maybe you've heard of it. It was basically a Hypercard approach to development, albeit with some additional technical tweaks, and an asset generation pipeline that was novel for the time.

Learn your computer game art history, or be doomed to repeat it. Remember that BAD art has occasionally made big money. Minecraft Alpha, Deer Hunter, Flappy Bird. There are reasons for that, most of them being about abandoning widely held industry assumptions, in the face of ongoing technical change.

If you're gonna try to sell "good" art, what's your strategy for anyone actually paying for it?

JoeJ wrote:

I'm afraid you still have a long way to go. To find out about yourself, who you really are...

Because something like a proud vibe coder can not exist. So you must be something else. ;D

Black holes, where the gravitational time dilation is extreme, are just black bodies that emit Hawking radiation. Also, the greater the kinematic time dilation, the closer a body is to becoming a black body. Internal processes inside of the body become interrupted… as in process interruption from the x64 architecture. What is the interrupting process? The black body process. It is viral, and doesn’t absolutely require gravitons. On this project I didn’t use much help from Claude.

taby wrote:

On this project I didn’t use much help from Claude.

But Claude is already monitoring your github. Once you figure out how gravity works, he'll claim the first Nobel Price given to AI.
At latest then you will learn AI does not benefit you, and you'll come back following the right path of using your own intelligence.

Maybe gravity is just the counter reaction to the expansion of the universe, trying to shrink it. \:?/

JoeJ wrote:

But Claude is already monitoring your github. Once you figure out how gravity works, he'll claim the first Nobel Price given to AI.

At latest then you will learn AI does not benefit you, and you'll come back following the right path of using your own intelligence.

Maybe gravity is just the counter reaction to the expansion of the universe, trying to shrink it. :?/

Well, if I ever win the Nobel, I won't be sharing my winnings with Anthropic (who are likely already rolling around in piles of cash). LOL

You're too set in your ways. Claude is indispensable. You can also use it to write a nice fluffy paper that can then be used as a skeleton for the final paper that you write.

Your guess is as good as mine. :)

So many fallacies here…

oneoffaccount said:
- Most people lack the funds to hire programmers to build prototypes for ideas they are unsure they should pursue.

Ok but most people aren't meant to be game developers. Not everyone is a plumber, not everyone is a carpenter. Or songwriters, or artists. Huge fallacy to advocate for universal ease of access for outsiders to something that should not be for everyone to begin with. All it does is flood the market with low-effort, low-quality works of people who aren't really meant to be in that field, pushing out the people who actually should be there. Real singers should be singers, I don't want AI-generated slop songs or autotune-distorted songs by people who can't sing.

oneoffaccount said:
I've changed my mind after seeing some of the attitudes openly displayed in a space meant to be fun and encouraging.

Making games is not all fun and games (pun intended). It's not good to encourage people in whatever delusion they are under, just so that everyone feels warm and fuzzy. At least not if you actually care about games and want to help people to make actually good games. Mentoring others means you share your expertise and you urge them to avoid making known mistakes.

oneoffaccount said:
- Sometimes people are bad with syntax or learning the quirks of a software language or debugging. Perhaps they have ADHD, Dyslexia, or other challenges. No, programming doesn't just "work" for all of them. Telling them to "just learn to code" is callous and able-ist and you should be ashamed of yourselves. Do you tell everyone to move their couches alone or "just run faster"? Do you rudely ask why everyone can't dance or dress equally well? Do you tell overweight people to stop taking transit? It would be very easy to ask pointed questions based on your behaviors, but that wouldn't be helpful here.

I am diagnosed with ADHD and I can program just fine, and I also saw lots of people with dyslexia program just fine, by hand. If you are mentally disabled to the point where you can't even problem solve problems that are required to be solved in programming, then sorry, but you simply can't be a programmer. And even AI can't help in that case, because it can't do the thinking for you. All it can do is the code writing, but you still have to do the planning. And if you can't think a plan through properly, then you're going to make something bad. Game design is hard and extremely complex. Those who aren't up for it due to disabilities, well, sorry, but someone should tell you that it's not realistic for you to do it.

I don't tell people to just run faster if they can't run faster, but I also don't tell someone in a wheelchair that they have a shot at running a marathon, and pathologically shield them from all criticism and feed into their vain delusion of becoming a runner. Wheelchair people should lead wheelchair lives. Which includes not hoping to become a runner.

And I'm not ashamed of giving people a reality-check.

oneoffaccount said:
If I catch an AI building something wrong, it tries to fix it. No quiet quitting. No overcharging of hours. No wild bureaucratic goose chase fiddling with the place and structure of my bug reports in an attempt to keep me from doing my bloody job in catching their errors.

So now you scoff at human programmers who get burned out at companies that don't treat them like people, and aren't motivated when they get overworked and underpaid, and constantly rushed with unrealistic deadlines? That's what causes quiet quitting. The fact that you praise AI for not quiet-quitting is quite cynical.

oneoffaccount said:
You should be encouraging people to dive into your field, and encouraging them to tackle challenges - not insinuate snidely and smugly that they're stupid or lazy for not being wired the same way you are, or for considering other approaches given their differing circumstances. Common anxiety issues aside - it is never as simple as "if you cared you would do it." Useless Gatekeepers.

Adios.

lol. Well, I'm glad you feel out of place here, because you just whine a long wall of text the moment someone wants to uphold a culture of excellence and high expectations. And I encourage people to dive into my field by “learning how to code” and I encourage people to actually “care enough to do it”. You spread a culture of mediocrity in favour of inclusion, in a place for enthusiasts, and you try to shame any enthusiasts who actually care about the field in question, and who are pained by mediocrity. But the thing is, people who actually love what they do will naturally strive for excellence. By defending mediocrity, you explicitly antagonise the people who truly care, and want to shoe-horn in people into the “community” that don't really care to begin with.

P.S.: I also highly recommend everyone to learn how to make their own graphics, 3D models, etc. — everything they need to make a game from scratch. If you have 0 clue what goes into making 2D sprite graphics, then how could you design a good 2D game, even if you direct someone else? If you have no clue about the efforts involved in various design choices, or the craft behind it, how can you even make an informed and deliberate decision? If you at least have the knowledge and understanding required to do it yourself in theory, and make a deliberate choice on what to achieve, then it's an entirely different matter whether you hire someone else or use some tool to speed up your work. But skipping all that and just prompting an AI without any understanding is terrible.

taby wrote:

I won't be sharing my winnings with Anthropic (who are likely already rolling around in piles of cash). LOL

You will. Soon the bubble will burst, and they will start to charge premium for AI.
If you are already addicted and dependent enough, you will pay.

oneoffaccount wrote:

Don't preach to me about what you think LLMs actually are - I already studied it a decade ago, long before I looked at Game Programming. It's extremely easy to iterate with AI for many obvious reasons. It's expensive and complicated to iterate with people, for equally obvious reasons.

It seems you lost contact to other humans and can't deal with them anymore, after spending so much time on LLMs.

Have fun iterating.

Proud humans don't iterate - they leap. \:D/

JoeJ wrote:

You will. Soon the bubble will burst, and they will start to charge premium for AI.
If you are already addicted and dependent enough, you will pay.

I admit that I pay for Claude. I've also had times of overuse, where they charge extra for extra usage. The results are chaotic... I mean, a little change in the prompt can lead to drastically different results. So, I try over and over to perfect the prompt, which produces a lot of usage. To be fair, they did give me a free month credit for no particular reason, other than maybe to pay me back for how much money I have spent. It's healthier than cigarettes. LOL

taby said:
To be fair, they did give me a free month credit for no particular reason, other than maybe to pay me back for how much money I have spent. It's healthier than cigarettes. LOL

At least they don't give you free samples of cigarettes, though. I personally prefer unaromatised cigarillos, meaning just pure tobacco, no weird paper or other substances such as synthetic aromas added. Never got addicted, nor did I get a necrotic lung or blood clots. Honestly, there's a much bigger risk factor for blood clots that everyone was hyped up about recently…

Also, oxidated (= burnt) nicotine is literally nicotinic acid, aka niacin, aka vitamin B3. I don't get how people still simultaneously believe in vitamins and also that the nicotine in smoked tobacco is somehow bad for you.

RmbRT wrote:

I don't tell people to just run faster if they can't run faster, but I also don't tell someone in a wheelchair that they have a shot at running a marathon, and pathologically shield them from all criticism and feed into their vain delusion of becoming a runner. Wheelchair people should lead wheelchair lives. Which includes not hoping to become a runner.

But... there are wheelchair marathons, so this metaphor breaks down quickly. I tell you, I know I couldn't win a wheelchair race, against people whose arms are conditioned to do it well.

The problem is none of us are really sure yet what makes AI assistance easier, the same, or harder. I've heard plenty of software development anecdotes of "harder" because of all the errors the AI makes, having to quality control that. It's all veering towards that unsatisfying engineering area of "it depends".

Which is why I suggested the OP ignore AI and figure out why their art is gonna sell anyways. It's not a new production problem.

Even mentally retarded people can use Scratch or something like it. I have no proof of this, but it follows from common sense — children, or the young of mind, are voracious learners. They are not disabled in this regard. As for schizophrenics, the disability practically does not exist, with the right medication, that is. I have lived experience with this disease.

to do anything meaningful in scratch, while you don't have to be able to type in syntax, you still have to do the mental equivalent. I don't think someone who is too mentally disabled to type in something in a computer is able to do all the other aspects of programming, which are way more complicated than remembering how to type in a formula or a for loop.

You could also imagine covering the full feature set of C++ in something graphical like scratch. The only difference between that and C++ would be that it would be much more tedious to use, but just as hard. All you end up doing is directly work on the abstract syntax tree instead of on the textual representation. Both are equivalent, and in both, you still have to be able to program and to “read” programs. You still have to be familiar with mathematical formulas and all that.

Someone who can write a pathfinding algorithm in a graphical mode is definitely also able to write it in text, even with dyslexia, even if it takes extra effort. And someone who can't write that in text, also won't be able to do it in an equivalent graphical editor. And if all you can do is use very high-level, domain-specific commands from a toddler-tier graphical language, then you have no business making games.