🛠️ Engineering Literacy
in the Age of Vibe Coding

🌊 The Vibe Coding Revolution

What changed — and why it matters to you

What is Vibe Coding?

• You describe what you want in plain English
• AI (ChatGPT, Claude, Copilot, Cursor…) writes the code
• You accept, run, iterate
• The barrier to writing code dropped to zero

The Opportunity 🚀

• PMs can prototype ideas in hours, not sprints
• Designers can build interactive mockups that actually run
• Solo makers can ship full products
• The gap between "I have an idea" and "here, try it" is vanishing

The Risk ⚠️

• AI code looks correct but may be subtly broken
• You can generate a mountain of code you don't understand
• No one reviews it → it ships → it breaks in production
• Speed without understanding = a ticking time bomb

🏗️ Why Engineering Literacy Matters

Speed without structure = tech debt on a credit card

Horror Stories 😱

The Pattern

Engineering literacy is the immune system that prevents this.

You don't need to be an engineer. You need to think like one.

📜 AI Instructions Files

Your engineering constitution for AI teammates

📜 Your Engineering Constitution

Write your engineering standards ONCE, AI follows them EVERY time.

What Goes in Your Instructions File

## Role
You are a senior engineering partner. I am a PM learning to code.
Explain what you're doing and why. Teach as you go.

## Code Style
- Descriptive names (getUserById, not doStuff)
- One function does one thing
- Prefer readability over cleverness

## Security
- NEVER hardcode secrets — use .env
- Validate all user input
- Use parameterized queries

## Testing
- Write tests alongside code
- Test happy path AND edge cases

## Git
- Meaningful commits: feat(auth): add login form
- Never commit .env or node_modules

💡 Add framework-specific overrides too — React patterns, iOS conventions, Python style sections.

Resources for AI Instructions

• 📚 github/awesome-copilot — community templates & examples
• 🔒 Snyk's vibe coding personalization guide — security-focused instructions
• 🌐 "One Prompt to Rule Them All" — cross-tool instructions pattern

🖥️ Bash & Terminal Fundamentals

The command line is your cockpit, not a scary relic

Why Terminal Matters

• AI tools output terminal commands — you need to run them
• GUIs hide what's happening; the terminal shows truth
• Servers don't have a desktop — terminal is all there is
• It's the universal interface to every dev tool

Essential Commands

# Navigate
pwd                    # Where am I?
ls -la                 # What's here? (with details)
cd my-project          # Go into a folder
cd ..                  # Go up one level

# Read files
cat README.md          # Print file contents
head -20 server.js     # First 20 lines
grep "error" log.txt   # Find "error" in a file

# Chain commands with pipes
cat log.txt | grep "ERROR" | wc -l   # Count errors

# Permissions & packages
chmod +x script.sh     # Make it runnable
npm install            # Install JS dependencies
brew install git       # Install git (macOS)

Reading Error Messages

$ npm start
Error: Cannot find module 'express'
    at Module._resolveFilename (node:internal/modules/cjs/loader:1075:15)
    at Module._load (node:internal/modules/cjs/loader:920:27)

• Line 1: What went wrong → Cannot find module 'express'
• Fix: npm install express
• 🚫 Don't just paste the error back to AI — read it first

🏛️ Code Architecture & Planning

Think before you prompt

The #1 Vibe Coding Mistake

"Build me a todo app"
→ 500 lines in one file
→ No structure
→ Can't add features
→ Start over

"Here's my file structure:
  /src
    /components
    /api
    /utils
  Build the TodoList component"
→ Clean, modular, extendable

Project Structure Patterns

my-app/
├── src/
│   ├── components/    # UI pieces (buttons, forms)
│   ├── pages/         # Full screens
│   ├── api/           # Backend calls
│   ├── utils/         # Helper functions
│   └── styles/        # CSS / design tokens
├── tests/             # Test files
├── public/            # Static assets
├── .env               # Secrets (NEVER commit)
├── .gitignore         # What git should ignore
├── package.json       # Dependencies
└── README.md          # What is this & how to run it

Key Principles

• Separation of concerns — each file does ONE thing
• Naming conventions — getUserById not doStuff
• Modularity — small Lego blocks, not one giant sculpture
• DRY — Don't Repeat Yourself (reuse, don't copy-paste)

📦 Version Control & Git

Your code's save game system

Why Git?

• 🔙 Undo anything — go back to any point in time
• 👥 Collaborate — multiple people, same codebase, no conflicts
• 📜 History — "who changed this and why?"
• 🛡️ Safety net — experiment without fear of breaking things

The Essential Commands

# Start tracking a project
git init
git add .
git commit -m "Initial commit: project setup"

# Daily workflow
git add .
git commit -m "Add user login form"
git push origin main

# Branching (safe experimentation)
git checkout -b feature/dark-mode
# ... make changes ...
git commit -m "Add dark mode toggle"
git push origin feature/dark-mode
# Then: open a Pull Request on GitHub

Commit Messages & .gitignore

node_modules/
.env
.DS_Store
dist/
*.log

🐛 Debugging & Troubleshooting

Read the error. Then ask AI for help.

The Anti-Pattern

Repeat until you have 47 open tabs and no idea what's happening

The Better Way

Debugging Toolkit

• console.log / print — the oldest trick, still the best

  console.log("user data:", user);  // What is this actually?
  console.log("reached step 3");    // Did we get here?

• Browser DevTools (F12) — Network tab, Console, Elements
• Stack traces — read bottom to top; your code is usually near the top

Common AI-Code Failures

• 🤖 Hallucinated imports — AI uses a library that doesn't exist
• 🤖 Wrong API version — trained on old docs, uses deprecated methods
• 🤖 Works in isolation — AI code works alone but conflicts with your existing code
• 🤖 Silent failures — no errors, but the logic is just… wrong

🧪 Testing

Your safety net against AI hallucinations

Why Test?

• AI-generated code is confident but not always correct
• Tests catch bugs before your users do
• Tests let you refactor fearlessly — change code, run tests, still works? Ship it.
• Tests are documentation — they show what the code is supposed to do

Types of Tests

Start with unit tests. They're fast, easy, and catch the most bugs per effort.

Test-Driven Prompting 🧠

Tell AI to write tests first, then the code:

Prompt: "Write unit tests for a function called
calculateDiscount(price, couponCode) that:
- Returns 10% off for code 'SAVE10'
- Returns 0 for invalid codes
- Throws an error for negative prices

Then implement the function to pass these tests."

• Tests define the spec — what "correct" means
• If AI hallucinates logic, the tests catch it
• You can read the tests even if you can't read the code

🔍 LLM Output Evaluation

Trust, but verify

Two Levels of Evaluation

Both matter. Level 1 is where you start. Level 2 is where you grow.

The Code Review Checklist

Before accepting any AI-generated code, ask:

• ✅ Does it actually do what I asked? (Not just look right)
• ✅ Can I explain what each section does?
• ✅ Did I run it? (Not just read it)
• ✅ Are the imports/libraries real? (Search npm/PyPI to verify)
• ✅ Are there any hardcoded secrets?
• ✅ Does it handle errors and edge cases?

Hallucination Red Flags 🚩

• Unknown packages: pip install super-easy-auth — Google it. Does it exist?
• Too-good-to-be-true APIs: magicLibrary.doEverything()
• Outdated syntax: AI trained on old data may use deprecated patterns
• Confident comments, wrong code: // This securely hashes the password … it doesn't

AI Product Evaluation Framework

• Rubrics: What does "good" look like? (accuracy 1-5, tone match, no hallucination)
• Eval Set: Representative inputs — normal cases + edge cases
• Scoring: Human judges, LLM-as-Judge, or programmatic metrics
• Iteration: Change prompt/model → re-run eval → compare results

Rubrics — Defining Quality

Rubrics turn "does it feel right?" into measurable scores.

Scoring Methods

Regression Testing for AI

• Changed a prompt or swapped models? Does quality go up or down?
• Run the SAME eval set before and after → compare scores
• Catches: model updates breaking your app, prompt tweaks with unexpected side effects

Tools: Braintrust, Promptfoo, LangSmith, OpenAI Evals

🔒 Security Fundamentals

Don't ship what you don't understand

AI's Security Blind Spots

Security Essentials

• .env files — secrets go here, never in code

  # .env (add to .gitignore!)
  DATABASE_URL=postgres://user:pass@host/db
  STRIPE_KEY=sk_live_abc123

• HTTPS everywhere — no exceptions
• Authentication — every route that touches data needs auth
• Input validation — never trust user input; sanitize everything

Security Prompt Pattern

When generating code, ALWAYS:
- Use environment variables for secrets (never hardcode)
- Add authentication checks to all API routes
- Sanitize and validate all user input
- Use parameterized queries (never string concatenation)
- Add rate limiting to public endpoints
- Set CORS headers appropriately

🔄 CI/CD Pipelines

If it's not automated, it's broken

What is CI/CD?

GitHub Actions — Your First Pipeline

# .github/workflows/ci.yml
name: CI
on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '20'
      - run: npm install
      - run: npm test

That's it. Every push now runs your tests automatically. If they fail, you know before deploying.

CD — Where Does Deployment Happen?

GitHub Actions — CI + CD Together

name: CI/CD
on:
  push:
    branches: [main]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
      - run: npm install
      - run: npm test

  deploy:
    needs: test
    runs-on: ubuntu-latest
    if: github.ref == 'refs/heads/main'
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
      - run: npm run build
      - uses: peaceiris/actions-gh-pages@v4
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}
          publish_dir: ./dist

• needs: test → deploy ONLY runs if tests pass
• if: github.ref == 'refs/heads/main' → only deploy from main branch, not feature branches

☁️ Cloud & Deployment

Package and ship your vibe-coded project

Docker — "Works on My Machine" Solved

# Dockerfile
FROM node:20-slim
WORKDIR /app
COPY package*.json ./
RUN npm install
COPY . .
EXPOSE 3000
CMD ["npm", "start"]

# Build and run
docker build -t my-app .
docker run -p 3000:3000 my-app

Where to Deploy

Start with Vercel or Railway. Graduate to AWS/Azure when you need to.

Deployment Checklist

• ✅ Environment variables set in production (not in code)
• ✅ HTTPS enabled
• ✅ Health check endpoint — can you tell if it's alive?
• ✅ Logging — can you see errors after deploy?
• ✅ Cost alerts — set a budget! AI-generated code can be resource-hungry
• ✅ Billing cap — don't wake up to a $10K bill

🎯 The New Skill Mix

You're the architect. AI is the builder.

What Changed

The Superpower Formula

All four layers = you can ship products independently

🗺️ Practical Learning Path

Concrete steps to build these skills

Week 1-2: Foundation

• 📺 Open your terminal. Run ls, cd, pwd. Navigate around.
• 📦 Install git. Create a repo. Make 5 commits.
• 🐙 Push to GitHub. You now have a portfolio.
• 📝 Create a simple project with the file structure from slide 4

Week 3-4: Build Something

• 🤖 Use Cursor or Claude to build a small web app
• 🧪 Ask AI to write tests first, then implementation
• 🔍 Review every line AI generates — look up what you don't understand
• 🐛 When it breaks (it will), debug before re-prompting

Week 5-6: Ship It

• 🐳 Write a Dockerfile for your project
• 🔒 Move all secrets to .env files
• 🔄 Set up a GitHub Actions CI pipeline
• 🚀 Deploy to Vercel or Railway
• 📊 Share the URL. You shipped. You're a builder now.

Resources

🎬 Takeaway

AI amplifies you.

Literacy determines what gets amplified.

The 10x PM is the one who can ship.

• You don't need a CS degree
• You don't need to write code from scratch
• You do need to understand what you're shipping
• You do need these 8 competencies as your foundation

Start this week. Open a terminal. Init a repo. Build something.

🛠️ The future belongs to builders.