Planning Before You Build
2.1: Why Planning Matters
The single biggest differentiator between successful vibecoding and frustrating vibecoding is what happens before you write any code.
Spending 2-3 hours planning saves 20-40 hours of confused building, scope creep, and rework.
WITHOUT PLANNING WITH PLANNING
Vague idea Crystal clear vision
-> Start coding -> Research landscape
-> Hit roadblocks -> Document insights
-> Pivot repeatedly -> Define requirements
-> Scope creep -> Plan execution
-> Burnout -> Build with confidence
-> Abandon -> Ship2.2: The Five-Step Pipeline
Every project flows through five steps before code is written:
The spark -- what you're building and why. One sentence for each. If you can't state it clearly, you're not ready to build.
Scan the landscape. What exists? What works? What tech stack fits? What are the constraints (budget, timeline, skill level)?
Your unique insights that AI doesn't have. Contrarian takes, domain knowledge, lessons from past failures, current market conditions. This is your competitive edge.
Concrete requirements. Goals, non-goals, user stories, technical specs, success metrics. This is the contract between you and your AI assistant.
Execution strategy. Tech stack decision, architecture overview, phased tasks, risk mitigation. This is the roadmap.
2.3: The Brainlift: Your Secret Weapon
The Brainlift is a document containing everything that AI cannot possibly know about your project:
- Contrarian insights -- what you know that contradicts conventional wisdom
- Domain knowledge -- industry-specific context from your experience
- Hard-won lessons -- what NOT to do based on past failures
- Current context -- recent market changes, competitor moves, things that happened after AI's training cutoff
# Brainlift Template
## Contrarian Insights
- [What do YOU know that contradicts the obvious approach?]
## Domain Knowledge
- [Industry-specific context AI won't have]
## Hard-Won Lessons
- [What NOT to do based on past experience]
## Current Context
- [Recent changes, competitor moves, new technology]AI generates based on patterns from its training data. Your unique perspective and recent information are what separate a generic AI output from a product that actually fits your market.
2.4: Writing a PRD
Your PRD turns vision into concrete, testable requirements. Here is the structure:
# PRD Template
## Overview
[Product name and one-paragraph description]
## Goals
1. [Primary goal]
2. [Secondary goal]
## Non-Goals (equally important)
- [What you are explicitly NOT building]
- [Scope boundaries]
## User Stories
- As a [user type], I want to [action] so that [benefit]
## Technical Requirements
- [Performance, security, integration requirements]
## Success Metrics
- [How you will measure success]
## Timeline
- Phase 1: [scope] -- [timeframe]
- Phase 2: [scope] -- [timeframe]Explicitly stating what you're NOT building is just as important as stating what you are. Without non-goals, every feature request feels like it belongs.
2.5: Writing a Gameplan
The gameplan is your execution roadmap. It answers "how will we build this, step by step?"
# Gameplan Template
## Tech Stack Decision
[Choice with brief justification]
## Architecture Overview
[High-level system design]
## Phase 1: Foundation
- [ ] Task 1
- [ ] Task 2
## Phase 2: Core Features
- [ ] Task 1
- [ ] Task 2
## Phase 3: Polish & Launch
- [ ] Task 1
## Risk Mitigation
- Risk: [potential issue] -> Mitigation: [how to handle]2.6: Feeding Your Plan to AI
Once you have your documents, this is how you start a build session:
Read docs/brainlift.md, docs/prd.md, and docs/gameplan.md.
Ask me any clarifying questions before we begin building.AI now has the full picture. It knows your goals, non-goals, constraints, architecture, and unique insights. The code it generates will actually fit your project.