- Joined
- Feb 8, 2021
This has been an infamous indie game dev brain worm for ages. I can't post my idea, someone will STEAL it and give me NO CREDIT while he makes MILLIONS (the idea is "Minecraft with the graphics of Skyrim")
Programming thread
- Thread starter Splendid
- Start date
-
🔎 If you see an alert for analytics.usips.org, this is a self-hosted system I am running as a test. It is GDPR compliant and collects no invasive information.
-
Want to keep track of this thread?
Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
Create account
- Joined
- Feb 15, 2024
It is more of a thing now at least.
As far as I am aware, the 'Needle in a Haystack' with a demo is the 'original' or at least the one that was public and tweeted about first. Lazy source.
- Joined
- Sep 10, 2021
I have no sympathy for you if you make a game that can be cloned by an AI straight away.
- Joined
- May 10, 2020
What kind of game would be hard to clone straight away? Even if you do something unusual, like performing pathfinding and collision detection on the GPU, a good LLM can still implement it for you.
- Joined
- Sep 10, 2021
Literally anything with artistic intent. We're talking about a game about sifting through a fucking pile of hay for a needle right now.
- Joined
- May 10, 2020
I don't think players care about your artistic intent, they only care if the game is fun to play.
- Joined
- Sep 11, 2024
Any game where more than few systems have to interact. All those AI games have "hard things" already done by game engines btw.
Games like that have been easy to make for years now. Any scriptkiddie could do it. Yet they still kept failing, being stuck in early access or demo stages.
The difference now is that you can do it with natural language over the most rudimentary understanding of programming.
- Joined
- May 10, 2020
Games where many systems interact are likely to be larger in scope and therefore less likely to be indie. Also, I don't think it's gonna stay this way forever, with how quickly the LLMs are improving.
- Joined
- Sep 11, 2024
I don't think there's that big of a correlation. AAA usually have more money thrown into scope, polishing and maybe performance.
Maybe, maybe not. My current bet is most of LLM hype is just gambling addiction.
I believe people underestimate how much time they waste tardwrangling LLMs. But because prompting is so easy compared to actually thinking, and seeing line count going up is rewarding in itself, it feels like there are those massive multiplicative gains.
Basically doom-scrolling but somewhat productive.
All that while you accumulate cognitive and tech debt.
Is it sustainable? I doubt it. But maybe the next model/harness/who knows what will be able to solve it.
名無し
kiwifarms.net
- Joined
- Dec 26, 2018
I've been working on a project recently that uses the image processing library vips. It purports to be a high level C library and even uses glib for reference counting. It also implements optional arguments in a funny way; most every function call ends with a NULL pointer because its arguments are variable length and null terminated. Each optional argument is also prepended with its name as a const char*. It ends up looking like this:
C:
vips_obj_func(obj, other_obj, arg, "option", value, "other_option", value, NULL);
SchizoDaemon
kiwifarms.net
- Joined
- Apr 17, 2026
Not this again. I can have an LLM scaffold a whole project in about 5 minutes. This would take me a good couple of hours.
Yes, you do have to tard wrangle them sometimes, but it's still much faster than manually doing stuff.
- Joined
- Sep 11, 2024
I am talking about iterating on already existing projects, not setting up fresh one that probably fits nicely in context and is quite similar to other projects out there.
Unimaginative Username
kiwifarms.net
- Joined
- May 27, 2025
AAA games are busy removing complexity and interacting subsystems to try to appeal to an ever-larger market segment in an attempt to recover their insane development costs. Dwarf Fortress, Kenshi, and Starsector are all indie passion projects and have much more interaction between systems/emergent gameplay than any AAA slop. Please try again.
- Joined
- May 10, 2020
Each of which took over a decade to develop, which goes contrary to the popular game dev advice of making small games. Cause in the same time frame you can manage to release 10 different games and there's a chance one of them will become a hit and cover the development costs of the ones that flopped, instead of betting all on a single game that can turn out to be a failure.
Cause game dev is still a business.
SchizoDaemon
kiwifarms.net
- Joined
- Apr 17, 2026
It is quicker on those as well. I've found LLMs much faster at doing things that I already know how to do, and while they are doing something, there has been massive improvement over the last year when I first tried doing this. I can also do other things while this is happening. Frequently the bottleneck isn't the LLM, it is the speed of the machine I am working with.
Last edited:
- Joined
- Sep 11, 2024
You say that based on your experience, I say it is not based on mine. What now?
There is this general trend in dev spaces of developer crying that they are now overworked because of performance expectations caused by LLMs. Which is interesting, because if they creating that massive gains I don't think it should be the case.
Previous studies on AI productivity gains showed that developers think they are more productive, but in reality they were less productive. Obviously those studies were done on older models, and measuring productivity is tricky anyway so we might dismiss it.
But these things while still anecdotal still point that LLMs might not be that great.
Anthropic itself released paper that showcases that using LLMs leads to impaired developer skill. A lot of productivity gains from AI require you to reduce reviewing code, and even if you do just briefly review you won't create internal model of your project as well as if you were solving problems yourself in the first place.
It's like tutorial hell but now with AI, it seems like you get it. But once something shit itself you are at disadvantage.
I might just be wrong, and coping. But I sill don't believe LLMs are that great for anything than oneshoting small scripts, or being used as semantic grep.
SchizoDaemon
kiwifarms.net
- Joined
- Apr 17, 2026
Have you considered I might be using it differently, or I might be using different tools?
Developers always cry about being overworked at these places. Some of it is valid, some of it is poor project management, and other times it is them having a moan. I've had to be the "hero" on a few projects and drag a whole project forward with nothing other than resentment, coffee, and a pack of peanut M&Ms. This was before AI.
Performance expectations/project management is a difficult problem that has been done poorly throughout the industry since its inception.
I've had no such effects. I am sure I am not as sharp as I was when I worked at <evil mega corp>, but that could be because I am working with morons and not being pushed to be better because expectations (and the developers) at <evil mega corp> were much higher than my current role.
I understand this point, but I can also have the LLM explain whatever piece of code or the process to me in plain English. I also have solved many of these problems myself many times before.
I am in my 40s, and my memory is nowhere near as good as it was when I was 25. When I was 25, I could be drinking in a bar and tell someone over the phone exactly what the relative path was for a file on a project with 1000s of files. I was a complete savant. I can no longer do this.
You are coping. I was coping with this until sometime last year. I've been able to use Qwen 3.8:27b (~20GB model) running on my GPU, and it can reasonably do all coding tasks on a large codebase, and it can handle it fine. Claude can do much better.
So when someone tells me that it can't be done or it isn't great for doing something, I am using it successfully to achieve exactly that. You must accept whatever you believe is probably at best innaccurate or incomplete.
- Joined
- Sep 11, 2024
I am not saying it can't be done. I am saying that performance gain is negligible and you pay hefty price.
And maybe you will have to accept that LLMs hijack your reward system the same way gambling does.
I guess time will tell.
SchizoDaemon
kiwifarms.net
- Joined
- Apr 17, 2026
You are asserting that the performance gain is negligible, yet I did a framework upgrade and migration in several hours that would have taken me several days. I am also not paying a lot of money. It is less money than my internet is.
What bullshit talking point is this? I used to work in the gambling business and used to hit the tables once a month after work. It isn't remotely the same. This is some bullshit you've read that sounds like it makes sense.
Go to a decent casino after being at the bar/restaurant in the sun. Go to the roulette/blackjack with a few friends or a chick with you, and lose some money for fun (doesn't have to be a lot) and come back to me then and tell me it's the same thing. There is a distinct ambience of being in a casino, at the Horses, the Dogs, or whatever. There is a massive buzz when the ball is bouncing around on the roulette table or when your horse race is neck and neck. Even online gambling, when you are invested in the game, has a certain rush. It is a whole experience.
Going to the bookings, playing the slots, or whatever online doesn't have any of the ambiance or a rush to it. Sure you will get gambling addicts down the bookies or on the slots machine down at the bar, but it is not remotely the same thing.
AI is nothing like that. I suppose it is cool when you can go, "Hey, can you do this?" and it does it, but that gets old after a while.
AI is much like regular development. I tell the computer to do something, I wait for a bit, see whether it works, review, and then iterate. That is what I was doing before. The difference is simply that I can have it do more. That is it.
The industry has already moved on. Most people have accepted that LLM usage will be part of their workflow.
Last edited:
- Joined
- Nov 21, 2020
I will say my slow local LLM doesn't save me real time, but it does save me effective time. I can say "Ok, take these 16 colors, make me a color selection box for these 3 colors in the code." Then wander off, mow the yard, whatever. And a few hours later come back and say "ok, that's great, but the border for the selected item needs to be more visible." wander off, come back and it's all good.
And I don't have to deal with cut-paste to build a GUI or any of that stuff. The highly timing critical stuff in the other project I did by hand because that's far more fun than UI work.
The only way you can tell the code was LLM generated is that it's actually working the way I want.
And I don't have to deal with cut-paste to build a GUI or any of that stuff. The highly timing critical stuff in the other project I did by hand because that's far more fun than UI work.
The only way you can tell the code was LLM generated is that it's actually working the way I want.