Flying Fire

Metroidvania solo developer. Working towards building mod tools to help anyone create worlds quickly and easily.

[Flying Fire @FlyingFireCo]

@UndeadMunch @Zalaphrax I can see that, I look at it from the solo dev perspective where it's one synergistic process. Often design decisions come from coding constraints for me

[Flying Fire @FlyingFireCo]

@7788FyhHH @ikar0x_dev @Zalaphrax Honestly ai coding in my opinion is an extremely beautiful art. Building graphs, loops, prioritizing specific gates and orchestrating subagents are so much fun and it's not as easy as "just ask and get what you want"

[Flying Fire @FlyingFireCo]

@7788FyhHH @ikar0x_dev @Zalaphrax I love ai coding, it's my favorite way to develop right now.

[Flying Fire @FlyingFireCo]

@7788FyhHH @ikar0x_dev @Zalaphrax I don't think it should, I'm a huge fan of using any resources available to achieve your vision. I do think that regardless of what tools you use the discrepancy of what you want to build and what you are able to accomplish is your personal style

[Flying Fire @FlyingFireCo]

@aozora_one @ikar0x_dev @Zalaphrax In this way I mean that all game code are game design ideas, but all game design ideas are not automatically code

[Flying Fire @FlyingFireCo]

@aozora_one @ikar0x_dev @Zalaphrax This is true, though it works both ways. It's not possible to always successfully implement your vision so the flaws of that translation are a big part of the artistic process

[Flying Fire @FlyingFireCo]

This is true, though a planned design is not the end product's design.

In game development, it's only possible to implement what is it possible to code which makes the end product's design fall into 3 categories:

1. What you are able to successfully implement exactly in line with your planned design
2. Trade-offs you make regarding something not technically possible to code given your knowledge and research
3. Accidental bugs that you wrote in code but ended up liking so you decided to keep even though it wasn't planned

For the end product these 3 come together to become the style of your game design, all only made possible by the limitations of the intersection of what coding allows and your ability and skill to write it in code.

It is in this that coding is in my opinion a fundamental part of game development.

Also I do not intend this post in an argumentative manor, just trying to relay my experience as a solo game developer

[Flying Fire @FlyingFireCo]

@ikar0x_dev @Zalaphrax Game design doesn't exist in a vacuum. It's literally the byproduct of code

[Flying Fire @FlyingFireCo]

@SurrealistShip When people specifically use it to separate models that output text from ones that output images, videos, or sound... yeah it's a big flag

[Flying Fire @FlyingFireCo]

we are so back lol - private branch on steam for Project Delta right now. going to validate on steam deck over the weekend. Potentially going public with the classic mode world next week. we'll see how playtesting goes. Fingers crossed no game breaking bugs

[Flying Fire @FlyingFireCo]

@legattacks @haylsho To me everything can be considered art. The distinguishing factor is that all art resonates with different people in different ways

[Flying Fire @FlyingFireCo]

I'll elaborate here with a specific example:

For my metroidvania i'm procedurally generating tilesets based on a room's collision info. This takes processing time. There are many ways to apply the rules of whether a specific 'top ground' or 'bottom ground' should go on a specific tile location, but we'll ignore that for the time being. Lets say we optimize that down so the algorithm does what we want in 500-2500ms.

When do we call the function that does that? Do we call it when the player enters the room? Then we have to make the player wait at least a half second for every room they enter looking at a black screen. We could apply the tileset graphics in parallel during runtime while inside the room. That could have a cool effect of 'digitizing' the graphics as each tile gets placed live - though that's a distinct style choice and probably only works the more 'cyber' your theme is for the game.

You could choose to pre-bake the aesthetics, but that requires storage space. If you have 1000 rooms and 30 different biomes that means 30,000 files with pre-baked data. This could get pretty large in storage and then you also need to consider an algorithm for reading the baked data and how long does that take and how to apply that? And then you also need a process to re-bake all the room data in the event any room or biome rules change.

Lets say you don't want to pre-bake and you want to generate on the fly, but you want to minimize load time since waiting a few seconds to load a room is getting really annoying during gameplay (I faced this issue working on my game). So you decide you'll kick off a parallel thread which processes the data for the current map during runtime and either caches it in memory or on disc (style choice). But then you must also consider how many processes should you use? Do you dynamically allocate based on the amount of cores the running client has available for greater speed? Or do you just do it serially one at a time on a single parallel thread?

There is no one best answer. There's a million questions and the point where you give up trying to research for a better solution. In my opinion the point in which you give up on trying to make your optimal vision true is your style. At least that's how I try to rationalize it.

[Flying Fire @FlyingFireCo]

@haylsho Optimized code is a style choice. It often comes with trade-offs. The is certainly not one best way to write a script especially with how it interacts with surrounding architecture.

[Flying Fire @FlyingFireCo]

I started with game maker then went to xna then unity. Xna was absolutely brutal but also taught me c# and also got me pretty close to the raw loop of game dev. When I discovered unity I was so happy to not have to manually write a draw loop or the main update loop and the engine just handled everything.

Unity is great for what it abstracts but is still a challenge to achieve a broad vision. I wonder if it's a "happy medium" where it's technical enough but challenging enough where people who started with engines are creatively liberated not being bogged down by engine code and able to focus on game design code so that they naturally fight against the next level of abstraction.

As someone who has been along for most of the ride - I personally have never been more excited by what ai enables. It really does feel like the same level step from xna to unity from unity to ai

[Flying Fire @FlyingFireCo]

Looking at a lot of these old raw texts i can't believe how much i improperly used "it's" vs "its" lol embarrassing. Still want to stay faithful to the original raw text, spelling and grammatical errors and all

[Flying Fire @FlyingFireCo]

[DSA Archives 7 - The Title Screen]
Written: August 2nd, 2018
Developed: August 17th, 2017

[Flying Fire @FlyingFireCo]

Original Text:

Day 6 of development. Now that the first level and boss was completed, it was time to start making the game feel like a game. I had a lot of fun making the screen where Flying Fire shows up. I wanted to make it feel just like turning on an original Game Boy (or Game Boy Color, which was what I had.) The scene just has the logo (which has now changed) and an object that's the same color as the background starting on top of it and moving to the right when the sound starts playing. It was pretty simple but I feel like it always adds a nice bit of professionalism to a game.

Then I spent a decent amount of time hand drawing out the title "Destroy Space Aliens." I think I used battle toad's logo as a reference image while doing so. Reference images is something that I'd reccomend to everyone who is doing any type of art. Don't copy right over the image, but it's nice to have the image to the side to help you get proportions right and have a little inspiration.

Programming the menu was pretty simple (other than having to develop a sprite font) I just had a list of objects that were text that the selected blip would move to whichever object was active at the time. Hitting up or down would change which one was active. Then when you pressed the start button, a timer would toggle on and off if the selected text and blip was visible while playing a bling. I also believe I hardcoded each response for the action of each field. I probably should have made each menu item inherit from a generic abstract menu class and each different object could implement it's own action to perform when clicking. But my hardcoded solution worked out well, and since I was the only one working on the project it probably saved time not making the code too complicated. In my own personal development process (which differs greatly than when I'm working with a team) I usually go with the fastest route to get the job done. If I need to fix it, I'll usually do that later.

Now that the game had a wrapper around it, it was starting to feel like a full product. This really helped motivate me to keep on working on the project, and reassured me that I'm at a point where I could develop this whole game. After 6 days I had some pretty good progress. However that WAS 6 days of non-stop working on DSA. While I can't remember exactly, I'm sure that on the 7th day of developing DSA I rested. Or went rock climbing or camping or something. I don't remember.

[Flying Fire @FlyingFireCo]

Imo "anti ai" was doomed to attract the worst kind of hatred in people from the beginning just because of the name. It's focusing on demonizing something instead of celebrating it. A "traditional arts" movement would be much more productive which celebrates and educates people how to do things the by hand.

And for ai itself, love it or hate it, in all domains in recent years it's become extremely transformative from regurgitating whatever has been put into it. The training process has become so advanced with synthetic data, rlhf, rlvr, etc that to call it stealing at this point is to deny the human creativity and engineering that has gone into creating the most transformative technology of our time.

As someone who spent my life learning things the old way, I don't want those practices to go away since I find immense value in that and hope others still learn that was as well in the future, but I'm also excited with how much ai has helped creatively liberate me and explore new difficult paths which I might have never explored before and honestly I don't really care what anyone else thinks - I make art for myself, not to appease others

[Flying Fire @FlyingFireCo]

I think this is where it falls apart. Ai is beginning to discover new math so by definition that's doing things that has never been done.

While yes it is generating tokens in the form of text, images, sound, and video - and indeed it is possible to calculate by hand or deterministic machine all possible permutations of words to form ideas, pixels to form images, wave forms to create sound, etc the permutations would take essentially infinitely long to find something useful

Ai models are the optimization of the search of that massive space. So in this way technically it's doing nothing new with regards to what's possible to output but it's a brand new vehicle for searching that previously impossible space