Core Philosophy

Core Philosophy

The Genie Mindset

You have a genie. It will build anything you ask β€” but only if you ask it correctly.

AI isn't a replacement for thinking. It's an amplifier for execution. You provide the vision, constraints, and context. AI provides the implementation speed.

YOU decide what to build (Eagle Sight: Idea β†’ Brainlift β†’ PRD β†’ Gameplan)
YOU decide how to structure it (Horizon: provide the context AI can't see)
YOU decide when it's done (Tornado: validate with Research + Logs + Tests)
AI helps you get there 10x faster

Stay in the Driver's Seat

The best vibecoding comes from clear direction, not vague requests:

You're DrivingYou're Drifting
"Add JWT auth using our existing User model, follow the pattern in AuthService.ts""Add authentication"
"Optimize this query to < 100ms, no schema changes allowed""Make this faster"
"Create checkout flow, here's my Brainlift with our payment constraints""Build an e-commerce site"

The Two-Phase Process

Elite vibecoding has two distinct phases:

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  PHASE 1: EAGLE SIGHT (Pre-Build)                           β”‚
β”‚                                                             β”‚
β”‚  πŸ’‘ Idea β†’ πŸ” Research β†’ 🧠 Brainlift β†’ πŸ“‹ PRD β†’ πŸ—ΊοΈ Gameplan β”‚
β”‚                                                             β”‚
β”‚  TIME: 2-3 hours                                            β”‚
β”‚  SAVES: 20-40 hours of confused building                    β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                            ↓
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  PHASE 2: THE 7-STEP PROCESS (Build)                        β”‚
β”‚                                                             β”‚
β”‚  1. Load User Rules (identity + guardrails)                 β”‚
β”‚  2. Index Codebase (architecture context)                   β”‚
β”‚  3. Read Specific Files (implementation details)            β”‚
β”‚  4. Research Docs + Online (external knowledge)             β”‚
β”‚  5. Write Code + Tests (implementation)                     β”‚
β”‚  6. Run Tests + Compilation (validation)                    β”‚
β”‚  7. Log β†’ Fix β†’ Repeat (feedback loop)                      β”‚
β”‚                                                             β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

The Three Principles

1. Oneshot Paradigm

When something breaks, don't fix forward. Go back to the original prompt with error logs.

❌ Prompt β†’ Broken β†’ Fix β†’ More broken β†’ Fix β†’ Chaos
βœ… Prompt β†’ Broken β†’ STOP β†’ Original prompt + error log β†’ Works

2. Tornado Building

Three forces spinning together solve any problem:

Research + Logs + Tests = Solution
    ↑_________↓_________↑
    Each feeds the others

3. Horizon Thinking

AI thinks vertically (top-to-bottom implementation). You must provide horizontal context (the landscape of your system).

AI sees:    "Make feature X"
You know:   Existing patterns, constraints, history, integrations
Your job:   Translate horizontal knowledge into the prompt

The 80/20 Split

AI handles ~80% of the work (boilerplate, patterns, translations).

You handle the ~20% that matters (architecture, security, business logic, context).

Know When to Type

Sometimes typing is faster than explaining:

  • Tiny changes β€” Rename a variable, fix a typo
  • Domain-specific logic β€” Only you understand the business rules
  • Flow state β€” Don't break your rhythm for a one-liner

Stay Curious

Use AI as a learning tool:

  • Ask AI to explain code you don't understand
  • Request alternative approaches to problems
  • Learn patterns from AI suggestions
  • Treat every interaction as an opportunity to grow

The Goal

The goal isn't to become dependent on AI.

It's to become 10x more effective by delegating the tedious parts and focusing your energy on what matters: building great products.


"The genie is ready. The question is: how well can you make your wish?"

Ready to learn pre-build clarity? Continue to Eagle Sight β†’