Framework
Follow along with the written framework below as you watch the curated video
What You'll Learn
- Where to run Claude Code (and where NOT to)
- How Claude "sees" your files
- Project structures that work WITH AI (not against it)
- The difference between dev projects and founder projects
Why This Matters (2 min read)
The Problem Most Founders Have:
You open Claude Code, point it at your Desktop folder, and ask "help me with my product spec."
Claude says: "I see 847 files. Which one is the product spec?"
You forgot you have 3 years of random downloads, old presentations, family photos, and tax documents in there.
Claude just tried to read everything. Your context window is now at 80%. You haven't even started.
The Real Issue:
Claude Code isn't ChatGPT. It doesn't live in a sandbox. It edits REAL files on your computer.
If you don't organize your projects properly:
- Claude wastes context reading junk
- You can't find what Claude created
- Files get lost in random folders
- You spend more time managing chaos than building
How Claude Code Sees Your Files
Claude Code works with three key concepts:
1. Working Directory (Where You Run Claude)
When you run Claude Code in VSCode, it starts in whatever folder you have open.
This folder is Claude's universe. It can see:
- Every file in that folder
- Every subfolder
- Everything in subfolders
It CANNOT see:
- Files outside this folder
- Your Desktop (unless you opened Desktop)
- Other projects
Example:
textYou open: /Users/sarah/my-projects/saas-idea/ Claude can see: ├── specs/ ├── research/ └── notes.md Claude CANNOT see: /Users/sarah/Downloads/ /Users/sarah/Desktop/taxes-2025.xlsx
2. File Tree (How Claude Navigates)
Claude reads your project structure like a map:
textsaas-idea/ ├── specs/ │ ├── product-spec-v1.md │ └── feature-list.md ├── research/ │ ├── competitors.md │ └── customer-interviews/ │ ├── interview-sarah.txt │ └── interview-alex.txt ├── marketing/ │ └── landing-page-copy.md └── README.md
When you say "Claude, read my product spec," Claude:
- Lists all files
- Searches for "spec" in filenames
- Finds
specs/product-spec-v1.md - Reads it
If your files are named randomly (Document1.txt, Untitled.md), Claude can't find anything.
3. Context Window (Claude's Memory Limit)
Claude Code has a 200,000 token limit (~150,000 words).
Every file Claude reads eats into this budget.
Bad structure = wasted context:
- Claude reads
node_modules/folder (10,000 files you don't need) - Claude reads old drafts you forgot to delete
- Context blown before doing real work
Good structure = efficient context:
- Only relevant files in working directory
- Clear naming (Claude finds files fast)
- Old stuff archived elsewhere
The Founder Project Structure (What Actually Works)
❌ What Founders Do (Wrong)
textDesktop/ ├── Final-Spec-V3-FINAL-USE-THIS.docx ├── competitor-research.txt ├── notes-random.txt ├── Screenshot 2025-01-15.png ├── taxes-2024.pdf └── family-vacation.jpg
Problems:
- Mixed personal + work files
- No clear categories
- Random naming
- Claude wastes context on family photos
✅ What Founders Should Do (Right)
Create a dedicated workspace for AI-assisted work:
text~/founder-workspace/ ├── product-validation/ # One project per idea │ ├── CLAUDE.md # Tells Claude about this project │ ├── specs/ │ │ ├── product-spec.md │ │ └── feature-list.md │ ├── research/ │ │ ├── competitors.md │ │ └── interviews/ │ │ ├── 001-sarah-interview.txt │ │ └── 002-alex-interview.txt │ ├── analysis/ # Claude outputs here │ │ └── (empty - Claude fills this) │ └── prompts/ # Your prompts library │ ├── 01-idea-validation.md │ └── 02-competitor-analysis.md │ └── marketing-campaigns/ # Another project ├── CLAUDE.md ├── landing-pages/ └── email-sequences/
Key Principles (Remember These)
1. One Folder Per Project
Don't mix:
- Product specs + marketing copy + customer research in one folder
Instead:
product-validation/(specs, research)marketing-q1-2025/(copy, campaigns)automation-workflows/(Zapier, scripts)
Why: Each project has different context needs. Keeps Claude focused.
2. Descriptive File Names
❌ Bad:
textDocument1.txt notes.md stuff.docx
✅ Good:
textproduct-spec-v2.md competitor-analysis-jan2025.md customer-interview-stephen.txt
Why: Claude uses filenames to decide what to read. "Document1" tells Claude nothing.
3. Clear Folder Categories
Use folders like:
specs/— Product requirements, feature listsresearch/— Customer interviews, market analysisanalysis/— Claude's outputs (syntheses, reports)prompts/— Your reusable prompt templatesdrafts/— Work in progressarchive/— Old versions (exclude from Claude's view)
Why: Claude navigates by folder structure. Categories = faster file discovery.
4. Keep Archives Separate
textproduct-validation/ ├── specs/ ← Claude reads this ├── research/ ← Claude reads this └── _archive/ ← Claude IGNORES this (starts with _) └── old-drafts/
Use _archive/ or .archive/ for old files.
Why: Claude skips folders starting with _ or . by convention.
The "CLAUDE.md" File (Your Secret Weapon)
Every project should have a CLAUDE.md file at the root.
What it does: Tells Claude about this specific project BEFORE you even ask anything.
Example CLAUDE.md for a product validation project:
markdown# Project: SaaS Idea Validation ## What This Project Is I'm a non-technical founder validating a B2B SaaS idea (team collaboration tool). This folder contains product specs, customer research, and competitive analysis. ## My Role - Founder & Product Lead - NOT a developer (don't write code) - Focus: Product specs, customer insights, go-to-market strategy ## What I Need Help With - Writing product specs from rough ideas - Analyzing customer feedback for patterns - Competitive research and positioning - Prioritizing features (RICE scoring) ## File Structure - `specs/` — Product requirement docs - `research/` — Customer interviews, feedback - `analysis/` — Your syntheses and reports (create here) - `prompts/` — My reusable prompt templates ## How to Help Me - Keep outputs concise (1-2 pages max) - Use founder-friendly language (not technical jargon) - Show your reasoning (don't just give answers) - Save outputs to `analysis/` folder with clear names
Why this works:
When you open Claude Code in this folder, Claude reads CLAUDE.md automatically.
You say: "Help me validate this feature idea: [idea]"
Claude already knows:
- You're a founder (not a dev)
- This is a B2B SaaS product
- You want concise, non-technical outputs
- Outputs go in
analysis/folder
No repeated explanations. Context saved.
Founder Project Templates (Download & Use)
We've created starter templates you can clone:
Template 1: Product Validation Kit
textproduct-validation-kit/ ├── CLAUDE.md (pre-configured) ├── specs/ ├── research/ │ ├── interviews/ │ └── competitors/ ├── analysis/ └── prompts/ ├── 01-idea-validation.md ├── 02-competitor-research.md └── 03-feature-prioritization.md
Use this for: Validating new product ideas, feature concepts
Template 2: Customer Research Kit
textcustomer-research-kit/ ├── CLAUDE.md ├── transcripts/ # Drop interview recordings here ├── feedback/ # User feedback snippets ├── analysis/ │ └── (Claude generates insights here) └── prompts/ ├── 01-interview-questions.md ├── 02-transcript-synthesis.md └── 03-pattern-finding.md
Use this for: Customer interviews, feedback analysis, persona building
Template 3: Content Generation Kit
textcontent-kit/ ├── CLAUDE.md ├── drafts/ # Your rough ideas ├── final/ # Claude's polished outputs ├── brand-assets/ # Tone guides, examples └── prompts/ ├── 01-landing-page-copy.md ├── 02-email-sequence.md └── 03-social-posts.md
Use this for: Landing pages, email campaigns, social content
Try This Now (Exercise — 10 Minutes)
Step 1: Create Your First Founder Project (3 min)
- Open your terminal or file explorer
- Create a new folder:
mkdir ~/founder-workspace/my-first-project - Open it in VSCode:
code ~/founder-workspace/my-first-project
Step 2: Add Structure (3 min)
Inside that folder, create:
textmy-first-project/ ├── CLAUDE.md ├── specs/ ├── research/ └── analysis/
Use VSCode's file explorer or terminal:
bashmkdir specs research analysis touch CLAUDE.md
Step 3: Write Your CLAUDE.md (4 min)
Copy this template and customize it:
markdown# Project: [Your Project Name] ## What This Project Is [1-2 sentences: What are you building or validating?] ## My Role - [Your role: Founder, PM, etc.] - [What you do: Product, strategy, etc.] - [What you DON'T do: Code, design, etc.] ## What I Need Help With - [Thing 1: e.g., Write product specs] - [Thing 2: e.g., Analyze customer feedback] - [Thing 3: e.g., Generate marketing copy] ## File Structure - `specs/` — [What goes here] - `research/` — [What goes here] - `analysis/` — [Claude outputs here]
Step 4: Test It
- Open Claude Code in VSCode (click the ✱ icon)
- Say: "Read my CLAUDE.md and tell me what you understand about this project"
- Claude should summarize your project context
If it works: You've successfully set up your first AI-ready project structure! 🎉
Common Mistakes (And How to Fix Them)
Mistake 1: Running Claude Code on Your Entire Desktop
Problem: Claude tries to read thousands of unrelated files.
Fix: Create a dedicated ~/founder-workspace/ directory. Only open PROJECT folders in VSCode.
Mistake 2: No CLAUDE.md File
Problem: You repeat "I'm a founder, I need help with product specs..." every session.
Fix: Write CLAUDE.md once. Claude remembers forever (per project).
Mistake 3: Random File Names
Problem: Files named Document1.txt, notes.md, stuff.docx
Fix: Use descriptive names: product-spec-v2.md, competitor-analysis-jan2025.md
Mistake 4: Mixing Projects
Problem: One folder with product specs + tax documents + family photos.
Fix: One project = one folder. Personal files stay OUT of work folders.
Pro Tips
Tip 1: Use Date Prefixes for Time-Series Files
textresearch/interviews/ ├── 2025-01-15-interview-sarah.txt ├── 2025-01-18-interview-alex.txt └── 2025-01-22-interview-ben.txt
Why: Easy to sort by date, Claude sees chronology.
Tip 2: Create a "Prompts Library" Folder
Save your best prompts as files:
textprompts/ ├── 01-idea-validation.md ├── 02-competitive-analysis.md └── 03-user-persona-builder.md
Why: Reuse prompts across projects. "Claude, run prompts/01-idea-validation.md on this new idea."
Tip 3: Use .claudeignore to Exclude Junk
Create a .claudeignore file (like .gitignore):
text# Don't read these _archive/ drafts/old-* *.pdf *.jpg *.png
Why: Claude skips these files entirely. Saves context.
Downloads
📥 Founder Project Structure Template.zip — Clone and customize 📥 CLAUDE.md Template (Founder Version) — Copy-paste ready 📥 .claudeignore Template — Exclude common junk files 📥 Project Setup Checklist (PDF) — Step-by-step guide
What's Next
Now that you know HOW to organize projects, let's talk about CONTEXT MANAGEMENT.
In the next lesson, we'll cover:
- What is Claude's "context window" (and why it matters)
- How to check context usage
- When to
/recapand start fresh - Managing large projects without blowing your budget
→ Next Lesson: Context Window Management
Key Takeaways
✅ Run Claude Code in dedicated project folders (not Desktop, not random locations)
✅ Use clear file names (product-spec.md, not Document1.txt)
✅ Create CLAUDE.md in every project (tells Claude your context)
✅ Organize by category (specs/, research/, analysis/)
✅ Archive old files in _archive/ or .archive/ folders
Sources & Attribution:
This lesson curates content from:
- Arun Iyer: Instruction Files for AI Coding Assistants
- Builder.io: Complete Guide to CLAUDE.md
- Anthropic: Using CLAUDE.MD Files
- Claude Code Official Docs: VS Code Setup
Additional insights and founder-specific adaptations by NerdSmith.
Version: 1.0 (2026-02-08) Module: 0 — Claude Code Bootcamp Lesson: 1 of 5
Downloadable Resources
Click any resource to download. All downloads are free for enrolled students.