Step 3: Initialize Memory Bank with AI

The Memory Bank is the brain of your project. It gives your AI assistant long-term context so it never "forgets" your decisions.

🚀 Quick Start

Get your AI's memory initialized in under 5 minutes.

1. Prerequisites

Required Files:

✅ Step 1: Product Description (product_description.md)
✅ Step 2: Features List (app_features_list.md)

The AI needs these files to generate accurate context.

2. Run the Generator

Copy the Prompt Navigate to ai_rules/prompts/3_memory_bank_init_prompt.md and copy the entire file.
Paste into Code AI Open your Code AI (Antigravity, Cursor, Claude Code etc.) and paste the prompt. It will read your product description and features list.

The AI analyzing your project context
Generate Files The AI will generate 9 markdown files. Create them in your project:
```
mkdir -p .memory
# Create files as instructed by the AI
```
The generated memory bank files in your project

3. Verification

Folder .memory/ created in project root
9 markdown files created inside .memory/
10-product.md contains your REUSE/CUSTOMIZE/NEW categorization
50-progress.md has today's date and "Initialization" entry

Memory Bank Sidebar Structure The .memory folder structure in your IDE sidebar

🧠 Deep Dive: How it Works

If you want to understand why this works and how to maintain it, read on.

The Problem: AI Amnesia

AI models have a limited "context window". As your project grows, they forget early decisions, hallucinate imports, and lose track of architecture.

U-AMOS (Universal AI Memory Operating System) solves this by creating a structured "external hard drive" for your AI.

The Structure: 3 Memory Tiers

The .memory folder is organized by update frequency:

Tier	Files	Purpose	Update Frequency
❄️ Cold	`00-description` `01-brief` `10-product`	Identity. The core vision and hard constraints.	Rarely
☀️ Warm	`20-system` `30-tech` `60-decisions` `70-knowledge`	Architecture. Patterns, tech stack, and lessons learned.	Monthly
🔥 Hot	`40-active` `50-progress`	Context. What we are doing right now.	Daily

File Breakdown

`00-description.md` & `01-brief.md`

Your elevator pitch and non-negotiables. The "Soul" of the project.

`10-product.md` (The Blueprint)

Contains the exact feature mapping you generated in Step 2.

REUSE: Paths to existing boilerplate code.
CUSTOMIZE: What needs to change.
NEW: What needs to be built.

`20-system.md` & `30-tech.md` (The Machine)

Documentation of your data models (Supabase), state management (Redux), and tech stack choices.

`40-active.md` (The Dashboard)

Crucial File. The AI reads this at the start of every session to know:

What feature we are building.
What we did last.
What we need to do next.

Markdown

## 🎯 Current Focus
**Goal**: Configure onboarding
**Active Files**: `src/features/onboarding/config.ts`

`50-progress.md` (The Journal)

A chronological log of major milestones. "We finished Auth on Jan 12th."

How the AI Uses This

When you start a session using the AI Startup Sequence, the AI reads these files.

Without Memory Bank:

You: "Add login." AI: Reinvents the wheel, ignores existing Auth system.

With Memory Bank:

You: "Add login." AI: Reads 10-product.md -> "I see we have an existing Auth feature at src/features/auth. I'll help you configure it."

Maintenance

The system is designed to be low-maintenance:

Daily: Update 40-active.md when you switch tasks.
Milestone: Update 50-progress.md when you finish a feature.
Ad-hoc: Update 70-knowledge.md if you solve a tricky bug.

What's Next?

Now that your AI has context, generate your visual identity.

Step 4: Design System

Generate your app's visual identity

Troubleshooting

Q: The AI says "I can't read files" A: That's normal for some LLMs. Simply copy-paste content from product_description.md or app_features_list.md if requested.

Q: "TBD" in memory files? A: Your input files (Step 1 & 2) were likely too vague. Regenerate them with more detail or manually fill in the TBDs.

Step 3 - Memory Bank