11-13-2025, 11:35 AM
Oh yeah, this isn't a money making plan for me, independent game development doesn't usually scream big-bucks. I've just always wanted to do it, and I can so I will. I had a game running really well years ago, the game engine was abandoned but I had a tone of fun. Always messed around with Unreal Engine. I do want to release something and give a good postmortem, that's really my big goals along with learning rust. (Also I'm renovating my house so my wife is onboard with me turning my basement into a tacky game dev studio for me and my two kids... don't judge me I got to run with this!)
I'm using AI mostly to learn rust with and stay organized. I've historically been a Java, C#, Javascript, Flutter/Dart dev focusing on web/device apps. So rust is a bit of mind bender but I feel like I'm making it further faster than me learning c++. Also one of my sons are learning and will be working on the project with me. I've tried to use AI a bit to speed things up, but it's so hit and miss that I've mostly given up. Adding logic later gets dangerous. I've had it start changing code all over because it 'thought' it's way was better even when I just tried using it to make a simple change to the name of a command. It tried to change the innovative plugin style command system I worked hard on to something more 'mainstream'. Then the nightmare of trying to debug what it does. I built a tcp test engine, I was lazy and tried to get AI to replicate it for my udp engine. I got a udp test that looked great, although formatting was an issue and it could not replicate my super simple text formatting. I looked at the data and noticed that it was too uniform. I asked AI if it was generated fake data, and of course... "Yes! You're absolutely right. To speed up the process it's just simulated data based on what you wanted. It's working perfectly!" I've found Gemini interesting in the sense that I had a few bugs I didn't understand because of not understanding rust. Gemini was by far the best to tell me what I did wrong and how to fix it and why it should be fixed like that. It's currently terrible at code though.
I've used godot for quite a while now. When I did try using AI with it I found chatgpt codex would actually get things more right then Claude code (this is all dependant on the week too, claude was really unusable for anything all summer). The only time I really let codex unleash on my client was with a 2d shader effect. It did it perfect in one go. I was surprised. It actually gave me a good template for similar effects also. Like I said, it's so hit and miss. I just manually work with godot since it's fairly easy, I just use gdscript because I haven't had a use for c# since my game is server heavy and it does add a bit of overhead. If I was building a godot single player or client heavy game I would definitely choose the .NET version.
I started with Bevy knowing that it's too early. The changes were just too numerous to keep up with and the documentation was quite a few versions behind and rarely updated. It was an easy switch to godot though.
I did play Kenshi but not very far, I should give it another go. I agree, there was a lot of ideas on how not to do things. I'll check out Kenshi 2 when/if it drops.
I got No Man's Sky for me and my two sons on a sale. For us, it's a mile wide and an inch deep kind of game (unless you just love exploring digital worlds). I think we were hoping more from the space combat, it was very lackluster to say the least. I've followed what I could on Light No Fire. Not much on it. I hope it will be good. They need to get away from the network model of No Man's Sky where you can just edit your saves and give yourself anything you want, it's client heavy but was just an after thought after people realized that there was no multiplayer when it first released. It also has issues syncing storms and weather. I'm sure they will address that this time, hopefully. Just worried it will be the same thing, great world gen but not much else.
Empyrion Galactic Survival was a fascinating idea but the dev didn't seem to care much about working on game AI, ever. Also destructible ground always leads to being able to burrow to win, wasn't fun in PvP.
I'm using AI mostly to learn rust with and stay organized. I've historically been a Java, C#, Javascript, Flutter/Dart dev focusing on web/device apps. So rust is a bit of mind bender but I feel like I'm making it further faster than me learning c++. Also one of my sons are learning and will be working on the project with me. I've tried to use AI a bit to speed things up, but it's so hit and miss that I've mostly given up. Adding logic later gets dangerous. I've had it start changing code all over because it 'thought' it's way was better even when I just tried using it to make a simple change to the name of a command. It tried to change the innovative plugin style command system I worked hard on to something more 'mainstream'. Then the nightmare of trying to debug what it does. I built a tcp test engine, I was lazy and tried to get AI to replicate it for my udp engine. I got a udp test that looked great, although formatting was an issue and it could not replicate my super simple text formatting. I looked at the data and noticed that it was too uniform. I asked AI if it was generated fake data, and of course... "Yes! You're absolutely right. To speed up the process it's just simulated data based on what you wanted. It's working perfectly!" I've found Gemini interesting in the sense that I had a few bugs I didn't understand because of not understanding rust. Gemini was by far the best to tell me what I did wrong and how to fix it and why it should be fixed like that. It's currently terrible at code though.
I've used godot for quite a while now. When I did try using AI with it I found chatgpt codex would actually get things more right then Claude code (this is all dependant on the week too, claude was really unusable for anything all summer). The only time I really let codex unleash on my client was with a 2d shader effect. It did it perfect in one go. I was surprised. It actually gave me a good template for similar effects also. Like I said, it's so hit and miss. I just manually work with godot since it's fairly easy, I just use gdscript because I haven't had a use for c# since my game is server heavy and it does add a bit of overhead. If I was building a godot single player or client heavy game I would definitely choose the .NET version.
I started with Bevy knowing that it's too early. The changes were just too numerous to keep up with and the documentation was quite a few versions behind and rarely updated. It was an easy switch to godot though.
I did play Kenshi but not very far, I should give it another go. I agree, there was a lot of ideas on how not to do things. I'll check out Kenshi 2 when/if it drops.
I got No Man's Sky for me and my two sons on a sale. For us, it's a mile wide and an inch deep kind of game (unless you just love exploring digital worlds). I think we were hoping more from the space combat, it was very lackluster to say the least. I've followed what I could on Light No Fire. Not much on it. I hope it will be good. They need to get away from the network model of No Man's Sky where you can just edit your saves and give yourself anything you want, it's client heavy but was just an after thought after people realized that there was no multiplayer when it first released. It also has issues syncing storms and weather. I'm sure they will address that this time, hopefully. Just worried it will be the same thing, great world gen but not much else.
Empyrion Galactic Survival was a fascinating idea but the dev didn't seem to care much about working on game AI, ever. Also destructible ground always leads to being able to burrow to win, wasn't fun in PvP.