My goal in building the Crazy Balls game was to validate the hypothesis: anyone, with no prior experience or outside help, can build, publish and sell a video game using agentic coding tools. Let's see what the data says.
Money
I set myself a budget before starting this experiment - I didn't want to spend more than $200 a month. I also tried to prioritise free tools as a first attempt at solving any problem, and only if those weren't enough, I was getting my wallet out.Here's my final costs breakdown:
- Codex $20 a month
- Canva $13 a month
- Game Engine - free (voluntary donation)
- Sounds and music - free (voluntary donation)
- Apple developer fee - $99
- Google developer fee - $25
- Microsoft developer fee - free
- Promotion & Marketing - $0
Time
My self-inflicted limitation was not to work more than 30 hours a week. And I mostly adhere to this. Practically, time split was the following:- Begging of dev: setting up tools = 30%, ideation and prototyping = 50%, testing 20%
- When scope was set: managing agents = 20%, preparing assets = 40%, testing = 40%
- When release version was ready: managing stores publishing = 80%, still testing = 20%
Effort
Working with coding agents is its own thing. I wasn't ready for how tiring it would be. Mostly because AI agents love long reads! Responses are long, commits are long, code solutions are long. Reading all of that takes a lot of time and effort. Understanding all of that is even more exhausting. I stopped trying to read code by the second week of development and later even gave up on reading commit messages. Instead, I used testing to judge the product state and quality. That might be viable for small products or games such as Crazy Balls, but it quickly becomes unattainable as the size of your product grows.Testing your product is another major effort you should be ready for. It starts with excitement (you're testing your own thing!), continues with mild irritation (you go through the same flows again and again) and ends with being absolutely fed up and needing a break (at one point I had auditory illusions - I was still hearing music from my game even when it was closed).
One more thing
It's a common trap in product development - endless improvement cycle. Most products are never done. There are always bugs to fix, new features to add, things to improve. The same temptation I had developing my game. With every playtesting session, every friend's feedback, every hour spent thinking about how the game could be improved - the scope just grew and grew. Until I stopped it. I reminded myself it was meant to be a small, simple experiment. So I made a very aggressive cut on the scope in order to publish the game sooner.It still pains me to see all the little and bigger snags I've left in the game, so many improvement areas, especially considering what I've learnt since I've published it... Yet, no more development! The promise made should be the promise kept.
Was it all fun and games?
Mostly, yes. Creating your own game is fun (especially if you're recovering from a corporate burnout). However, creating a publishable product is still lots of work, and it is still tiring. Especially when you're doing it alone. Product development is a team sport for a reason. Combined creativity, experiences and efforts of multiple people will always beat solo creators.Does that mean I am stopping? Not even for a second :) Tune in two weeks to see what's next for my little game-building adventure.
