# GameKree8r: AI-Readable Project Brief

Canonical project page: https://markvizion.com/projects/gamekree8r-engine
Live application: https://gamekree8r.netlify.app/
Owner: Mark Barnes / MarkVizion
Category: ai
Stack: Python, Ollama, HTML5 Canvas, Multi-Model, Next.js

## Summary

An autonomous game factory powered by multi-model AI agents that research, design, code, test, and publish HTML5 games — zero human code written.

## Problem

AI can generate code snippets, but shipping playable games requires reusable constraints, asset decisions, scoring loops, and quality gates.

## Architecture

- Multi-agent pipeline separates design reasoning, code generation, test/play loops, and publishing.
- Small HTML5 game SDK limits the code surface the model must generate.
- Arcade frontend turns generated outputs into playable public proof.

## Evidence

- Live hosted application: https://gamekree8r.netlify.app/
- Live hosted game factory and arcade surface.
- Five published games reported at 10/10 quality with zero fix attempts.
- Demonstrates agent orchestration, SDK design, and shipped interactive outputs.

## Tradeoffs

- Constrained genre templates improve reliability more than open-ended game generation.
- The SDK is the leverage point: it lets models focus on game logic instead of boilerplate.

## Outcome

A working AI game factory that proves Mark can design systems where AI repeatedly ships usable artifacts.

## Authority Signal

Strong evidence for autonomous workflow systems, AI production loops, and creative code generation.

## Keywords

Game Factory, Multi-Agent AI, Zero Human Code, 10/10 Quality, Python, Ollama, HTML5 Canvas, Multi-Model, Next.js, Strong evidence for autonomous workflow systems, AI production loops, and creative code generation.