Cover

Dino Clash

Inspiration

Dino Clash combines two gameplay loops we love: the relaxing progression of raising dinosaurs and the fast-paced strategy of merge-defense auto battlers.

Our goal was to make the ranch more than a collection system. The dinosaurs you raise, feed, and evolve directly become your battle team, making progression on the ranch meaningful in every fight.

We also wanted to explore whether a complete 3D game could be built with an AI agent driving the Cocos Creator editor, not just generating code.

What It Does

Dino Clash has two connected parts:

Ranch

Explore a small open world while riding your dinosaur.
Build hatchery pens and raise new dinosaur species.
Buy specific eggs with no gacha.
Feed and evolve dinosaurs to increase their levels.

Battle

Build your army by placing dinosaur nests on a limited grid.
Merge matching nests to create stronger units.
Manage limited space and resources to optimize your setup.
Choose your army's HOLD LINE and defend against enemy waves.
Choose passive upgrades during battle.
Survive a series of increasingly difficult stages.

The core loop is simple: raise dinosaurs → build your battle setup → fight → improve your collection → fight again.

How We Built It

Dino Clash was built with Cocos Creator 3.8.8, WebGL, and TypeScript, with a portrait 720×1280 layout.

The project was developed with Claude Code driving the Cocos Creator editor through an MCP bridge across 34 development sessions.

We used an editor-authored architecture, data-driven configuration, a pure battle simulation, automated validation gates, and real browser-based gameplay testing to keep the AI-generated codebase reliable.

Challenges & Learnings

The biggest challenge was making sure the game actually matched the intended design while working with an AI-driven development process.

We learned to:

Verify the real build instead of relying on editor previews.
Turn recurring bugs into automated validation rules.
Measure gameplay problems instead of relying only on subjective feedback.
Test complete player flows rather than individual functions.
Keep a build log so each AI session could build on the previous one.

Accomplishments

Built 2 playable scenes and 9 screens.
Created a complete ranch-to-battle gameplay loop.
Implemented dinosaur raising, evolution, spawning, merging, passive upgrades, and wave-based combat.
Built a pure battle simulation for automated balance testing.
Added automated code, configuration, UI, and gameplay validation.
Improved combat positioning based on measurable gameplay data.
Developed the project through 65 AI-assisted development sessions.

What's Next

Our next priorities are to rebalance merging so it becomes a meaningful strategic choice, create a proper stage progression with unique waves and bosses, expand the dinosaur roster and exploration maps, and replace placeholder assets with production-ready content.