# Building Crownlands 100% With AI: What I’ve Learned Along the Way — retained record

> Retained reddit record from reddit:gamedev.

Canonical: https://highsignal.app/data/records/2e9fec1bd35d3dcd

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:gamedev)  published 2026-10-04 11:10 ingested 2026-10-05 02:37

[original source ↗](https://www.reddit.com/r/gamedev/comments/1wxd42m/building_crownlands_100_with_ai_what_ive_learned/)

## retained text



```
I’m building **Crownlands, a medieval multiplayer strategy game, 100% with AI**, and I wanted to share what that journey actually looks like.

I bring the ideas, make the decisions, test the results, and keep pushing toward the game I want to create. AI helps turn that direction into code, artwork, interfaces, and working systems. The amount of discussion, testing, and revision behind each feature has been one of the biggest lessons.

A request like “let players attack another city” sounds simple. Then you start asking what happens if reinforcements arrive during the march, the defender changes clans, someone loses connection, or the attacker changes equipment before the battle finishes. Every answer affects another part of the game.

That’s where a lot of the real work happens.

Our creation process has gradually become more structured:

1. **Define the idea.** What should the player experience, and how does it fit the rest of the game?

2. **Work through the rules.** Discuss the details, edge cases, and possible problems before building.

3. **Build a focused change.** Keep the update small enough to understand and review.

4. **Test and question it.** Check the visuals, gameplay, multiplayer behavior, and anything else the change could affect.

5. **Refine and verify.** Fix what we find, review the update, and check the published version after release.

One of the biggest things I’ve learned is that **AI needs clear direction and a reliable record of decisions**. We now maintain a master specification for the game’s rules and an art guide for its visual identity. Those give us something consistent to return to as the project grows.

The artwork has taught me that too. A beautiful image still needs to fit the world, remain readable at game size, and work alongside everything around it. We’ve spent time defining Crownlands through parchment, worn stone, timber, banners, and a grounded medieval style.

There have also been plenty of less glamorous lessons: missing reports, reconnect problems, crowded maps slowing down, and updates needing more testing than expected. A feature can look finished on screen while still needing work underneath.

I’ve learned to question confident answers, ask for evidence, and separate “the code is written” from “the player can use it reliably.”

AI has made it possible for me to keep turning ideas into something playable. It has also pushed me to learn about game design, balance, performance, testing, and how to explain exactly what I’m trying to achieve.

Crownlands is still a work in progress. I’m proud of what we’re creating, and I want to share the fixes, mistakes, and decisions along with the screenshots. They’re a big part of how this game is being made.

  

ASK ME anything related; I might not know much, but I know some :)
```



document fetched 2026-10-05 00:25 · canonical https://reddit.com/r/gamedev/comments/1wxd42m/building_crownlands_100_with_ai_what_ive_learned
