My First Time Working with UE + AI on This Scale
I’d like to share a little bit about the original plan for my game, what didn’t work out, and what I’ve learned along the way.
The first thing you need to know is that I wasted an entire day. I tried to do two things at once: create a zombie survival game and integrate a second AI bot into UE 5.8.
As a result, I lost the whole day and decided to make something less ambitious than a full-blown survival game with crafting, biomes, and all that stuff. You know what I mean. It just wasn’t going to work.
I remember enjoying playing DarkOrbit as a kid. The problem is, it’s not the same game it used to be.
When I started looking at the theme, the idea of a Director’s Cut came to mind.
But here’s the thing: I’m the director, and this is going to be my version of that story.
- A Rough Start
At first, it was really difficult to understand how everything worked and get things under control. Remember, I was already a whole day behind schedule and had absolutely nothing ready.
At some point, I decided to pull myself together, focus, and actually get something done.
Pretty quickly, I managed to put together a prototype. It didn’t look particularly good, but at least it worked.
And instead of triangles, I finally had my first 3D spaceship.
- Planning Problems
I had more plans and ideas than I could realistically implement. The problem was that I hadn’t fully understood how many of the tools worked.
But I had an idea: wherever one AI agent wasn’t suitable for the job, I’d bring in another one.
I created a sub-agent responsible for maintaining documentation, while also making the main agent do the same.
Why?
One agent was responsible for translating things and pushing updates to Git. The other was the main agent, handling the actual development. That’s why having proper documentation was essential for its work.
Both agents needed to stay on the same page, and documentation was the key to making that happen.
AI still struggles with some .gitignore files, but let’s not get into that.
For now, let’s just call it a win. My first real success in a long time.
- Understanding and Getting Things Back on Track
The most overwhelming thing about finally understanding AI tools is realizing just how many possibilities they open up.
Suddenly, I can generate 3D models that would have taken me five years to create on my own. It improves code quality and can even generate artwork — something I’ve always struggled with.
But hey, AI is AI. It can handle things pretty quickly.
Need a green spaceship? Here you go!
Bigger cannons? Coming right up!
**AI doesn’t wait. AI delivers.
**
- There’s Always Something That Goes Wrong
Unfortunately, as always, something had to go wrong.
First, I ran out of my free Claude Code credits. And for some reason, that sent me into a bit of a downward spiral…
(I don’t know what to do with this now)
The biggest loser in this whole situation, including that first wasted day, was me and my ideas.
I had quite a lot of ideas, but unfortunately, bringing them to life would have required much more time.
Originally, I wanted players to be able to purchase their own space station and expand it by adding different modules. Sadly, that plan had to be scrapped due to time constraints.
Even getting a non-functional prototype up and running took me five hours.
Unfortunately, I had to make a choice: either finish the game or finish the space station system.
- So, Am I Happy with the Final Result?
I’d say… somewhat.
Unfortunately, due to time constraints, I had to lower my standards in some areas, and it will show.
There are no proper engine flames. The UI might be decent, but it’s nowhere near what I originally wanted.
One thing is certain, though: this project has given me some motivation to keep going.
- Final Thoughts (For Now)
We’re almost there.
I think that if a model like Fable 5.2 becomes available on the Pro plan without burning through usage limits as quickly as it does now, game development might never be the same again.
Right now, one of the things holding AI back is the relatively high cost of running these models compared to the time they save. But if that cost goes down and better models become more affordable, then…
There probably won’t be a game out there that doesn’t use AI in some way.
Well, unless it’s made purely as a hobby project.
- What’s Next for AI, Me, and Game Development?
I want to keep making games. But will I keep using AI?
If I somehow manage to get my hands on an amazing subscription plan, then probably yes.
Unfortunately, AI is currently too expensive a tool for me. It’s incredibly useful, but still too expensive.
I think that’s the biggest problem with AI right now. The technology is good, but the pricing hasn’t caught up yet.
If I had access to Fable 5.2 at the same price as Claude Pro, I think a lot of things would change.
For now, you have to work with what you’ve got. Even free plans can sometimes be enough to get the job done.
Thanks for reading!






