The CIGDI Framework: An AI-Powered Workflow for Indie Devs

Version: v 0.2 (WIP)

Last Update: 31/Oct/2025, 00:08:32

Build better prototypes, faster. This page introduces the CIGDI Framework, an open-source workflow and Unity project designed to help indie developers work smarter. Using a Co-Intelligence approach to AI, this framework streamlines the chaotic early stages of development so you can focus on what's important: finding the fun.

Framework: CIGDI in Practice

After three months, we reverse-engineered our survival tactics into a framework. Seven stages, two decision gates:

• 00. Research (AI-Assisted): Market analysis, player research → Research Documents
• 01. Ideation (AI-Enhanced): Brainstorming with AI as thought partner → Concepts Backlog
• [DECISION GATE 1: Priority Criteria] ← Human evaluates against capacity
• 02. Prototyping (AI-Supported): Build with AI technical help → Playable Prototype
• 03. Playtesting (AI-Analyzed): AI processes feedback at scale → Playtest Reports
• 04. Review (AI-Generated): AI drafts analysis, humans interpret → Documentation
• 05. Action (Human-Led): Humans make all strategic decisions → Updated Priorities
• [DECISION GATE 2: Timeboxing] ← Human decides: ship, cut, or iterate
• 06. Integration (Human-Led): Combine approved prototypes → MVP

The Golden Rules

Rule 1: AI Is Your Production Assistant, Not Your Creative Director

DO use AI for:

• ✅ Code scaffolding and boilerplate
• ✅ Documentation drafting
• ✅ Explaining technical concepts
• ✅ Debugging help
• ✅ Research synthesis

DON'T use AI for:

• ❌ Core game design decisions
• ❌ Art direction (unless you're doing AI art intentionally)
• ❌ Music composition (unless that's your explicit choice)
• ❌ Narrative design choices
• ❌ Any decision that defines what makes YOUR game unique

Why? AI optimizes for "generic good." Your game needs "specific right."

Rule 2: Understand Before Implementing (Trust But Verify)

If AI gives you code you can't explain, don't use it yet. Seriously.

The 5-Minute Test:

1. Can you explain this code to a friend?
2. Can you modify it if requirements change?
3. Can you debug it when it breaks?

If you answered "no" to any of these, you have comprehension debt. That's when you have working code you don't understand. It WILL come back to bite you.

What to do instead:

• Ask AI to explain the code
• Look up the patterns/concepts it's using
• Try implementing a simpler version yourself first
• THEN use the AI version if you understand it

Rule 3: Build Decision Gates Into Your Process

AI will generate infinite ideas. You need to filter them ruthlessly.