This post is written entirely by me (moi). No “ChatGPT wrote my homework” this time — though I’ll admit I often run my drafts through a quick tidy pass after. Later note : The image is 100% AI.
Decisions, Decisions
There’s a lot to decide when you’re building a strategy game — and I’m not a game developer, not a software developer, and definitely not a maths wizard.
So when we’re working through the Paper Rules (by “we”, I mean ChatGPT and me) and the rules ask things like:
- How do satellites work?
- What’s the sensor range?
- Do multiple satellites stack — and if so, how?
- Is it a flat bonus, a percentage bonus, or diminishing returns?
I hit an immediate problem: I can’t responsibly pick “sensor range = X” if I don’t even know what X means yet. How far apart are stars? How big is the map? What units are we measuring in?
Iteration (and why I’m leaning on it hard)
None of those questions are impossible — but they do require iteration.
Iteration is basically: make a best-guess decision, lock it in, move forward… and then adjust later when real context shows you what’s broken.
That sounds long-winded (it is), but it’s still better than getting stuck in an endless branching maze of “what ifs” and never building anything real.
And no — I’m not just spraying random numbers everywhere. A lot of the time sink is me trying to reason through how a system should behave, then picking numbers that feel plausible enough to test. Sometimes I get it right. Sometimes a planet dies before turn 1 finishes processing. When that happens, I iterate and try again.
v0.1 safe defaults (aka: build the machine first)
One thing I’ve learned fast: when something is unknown, you don’t pick a “perfectly balanced” value — you pick a v0.1 safe default. That means a default that’s implementable and deterministic, even if it isn’t “finished design”.
Or put another way: I’m choosing numbers that let me build the machine first — tuning comes later.
A real example: we changed how long a turn is
We originally set 1 turn = 1 month. On paper, that meant ~120 turns for 10 years — but we weren’t seeing meaningful population and economic change in that timescale (at least not in a way that felt satisfying for the kind of game I’m aiming for).
So after a couple of Paper Rules versions with that “1 month per turn” idea locked in, we iterated: we now use 1 turn = 3 months. That brings it to ~40 turns for 10 years, which makes long-term growth and strategic planning much easier to feel.
This is a “long haul” kind of game, and I’m really hoping players enjoy that slow-burn development.
The bigger question I’m actually testing
I know this approach won’t be for everyone — but it’s central to the experiment:
Can a total “Joe Nobody” (no coding experience) build a fully fleshed, releasable strategy game with the aid of AI… without clicking a magical “make my game” button?