The reality of AI in Iron Legacy

My technical background

I think I mentioned it somewhere already, but I am not a developer capable of writing high-quality code that is needed for game development. I consider myself more of a game designer or director in the case of developing the Iron Legacy (I have multiple roles as I am handling everything alone, but with an awesome support from my family and friends >> battletesting every build I made!! :D). My real profession is something even more different, but it sure comes from the IT world.

I always dreamt about making my own game, but as I didn't have enough time and skill to do that without the AI, so it never happened in the past, until now :)

The AI

I think that today's world is somehow spoiled or blurred by AI, and AI has become a buzzword that is overused in everything from home appliances to software products in general. This means that not everyone is capable of imagining what it really is to use the AI for any sort of development. It varies from the most basic tasks to more complex ones, and is also based on your style and usage of such an AI. In my case, I am using an IDE (the thing where you write the code) with an integrated AI agent, and I can also choose the model of the agents, the less capable ones, but cheap, or the more capable ones, but also more expensive.

The game has not been born by a single promt .. don't imagine it as me telling the agent .. "hey .. create some game for me and come with a good name so people like it and buy it, so I am rich..." (:D) This is the exact opposite of how development with AI looks.

Steam post image(btw the watermark could have been hidden from this image, but I don't want to, as I am using an AI for generating art images :) ) I am using an agile methodology, which is heavily focused on UX (user experience) and early prototyping. When I was playing with the idea of creating my own game in my head, I (naturally) went and discussed it with the AI (:D), just the regular one, the chat, we exchanged few ideas and after 5 to 7 "prompts" (they were pretty huge :D) I was sure enough to take it to the next level. I opened the IDE and created a new project with the name "The Game" (it is still named like that :D)

I have extracted the ideas that were shaped with the discussion with the previous AI, and transferred them to the more capable AI in the IDE. Btw, the discussion was completely driven by me shooting ideas and sharing my vision of the game with prompts to give me critical feedback from the game designer perspective = I was using the first AI as someone who could serve me as a good enough discuss partner with critical thinking. We created the game design document later in the IDE and few hours later (I was really focused on the discussion and designing the core game principles with the goal of:

1. not forgetting anything that came to my mind during the brainstorming
2. and trying to define the minimum viable product, or better, some sort of proof of concept

Once we had the GDD "ready" and the PoC defined, the agent was eager to start implementing... I was trying something similar, like 1 or 2 years ago, and it was pain ... I was stuck in copy pasting code from one window to another and trying to run the broken environment ... it was hell for me .. but now ... this is where the story begins :D The agent just started typing the code, and I was like ... wait .... is it already creating something or what...?! And the agent was like .. there it is ... I am done .. try it ... so I tried it and .... it worked :D but it was an extremely simple prototype of the game :D There was just a black background and two rectangles. The blue one was me :D and the red one was the enemy... I was able to move the blue one (me), and I was able to come to the red one, and one of us disappeared when we were close together for some time ... this was the first visualisation of the combat btw ... and all of this took like 5+ hours .... :D which is good on the one side but also not so good on the other side...

I can go deeper if you want to :) and share with you what the developing with the AI looks like in much deeper detail.

The outcome

I wanted to say that you, as a human "developer", are still in charge of everything. You can imagine the AI as your brutal workforce tool, but it is only doing things you tell it to do. It is not a miraculous worker that does everything for you.

Right now, I am redesigning the "widget" which shows you if it is a day or night in the game. I had to create visuals, and I did create some low-fidelity designs of how I am imagining the widget. Then it took like 20 prompts to get the widget to its final form that I have imagined.

The most funny iteration was on picture three, when I told the AI to place the widget above the map ... I wasn't specific enough obviously :D Later... when I was being hopeless, I used the low fidelity design once again to pinpoint the differences, but it was like playing the game of find 5 differences ... Steam post image

This is the final result btw :) which I am pretty happy with, but just this widget rework took me like 3 hours. Yes, I was able to rework all widgets into a single component that is now used globally for this widget, plus a bunch of other stuff, but the majority of the work was orchestrating everything, deciding about everything and checking the code, the results and then battle testing everything in the game and fixing unforeseen errors. Steam post image

I hope you all appreciate my honesty, and if you want to, I can tell you the rest of the story of how the Iron Legacy was born (read: Forged in a cosmic fire on a cosmic anvil by a god :D) in the future posts :)