ai-terminal-course

3 min read Original article ↗

AI Terminal Course

The first terminal-only AI engineering course. No videos. No slides. Just you, the terminal, and real API calls.

What is this?

Learn AI engineering by doing, not reading. Type every command. See real responses. Build production patterns.

Think vimtutor, but for AI engineering.

Quick Start

# Try it instantly (no install):
npx ai-terminal-course

# Or install globally:
npm install -g ai-terminal-course
ai-terminal

First run: You'll be prompted to add your OpenAI API key.

# Setup (one-time):
mkdir -p ~/.ai-terminal-course
echo "OPENAI_API_KEY=sk-your-key" > ~/.ai-terminal-course/.env

Get your API key: https://platform.openai.com/api-keys

What You'll Learn

20 modules. 30 min each. $10 total cost.

Week 1: Foundations

  1. Tokens - How LLMs see text
  2. LLM Calls - Your first API call
  3. Streaming - Real-time responses
  4. Embeddings - Semantic search
  5. Context & Memory - Handling conversations
  6. Structured Outputs - JSON, validation

Week 2: Production Patterns

  1. Tool Calling - Let LLMs use functions
  2. Prompt Engineering - Write effective prompts
  3. RAG Basics - Retrieval augmented generation
  4. Chunking - Document processing
  5. Observability - Monitoring with Helicone/Portkey
  6. Cost Optimization - Caching, model selection

Week 3: Advanced

  1. Evals & Testing - Validate AI outputs
  2. Agentic Patterns - ReAct, planning, reflection
  3. Context Engineering - Context window optimization
  4. LLM Gateways - Fallbacks, routing, retries
  5. Guardrails & Safety - PII detection, content moderation

Week 4: Ship It

  1. Semantic Caching - Similarity-based caching
  2. Production Monitoring - Logs, traces, costs
  3. Capstone Project - Ship your first AI feature

Who This Is For

  • ✅ You can code (any language)
  • ✅ You're comfortable with the terminal
  • ✅ You have ZERO AI knowledge (perfect!)

Why Terminal-Only?

No distractions. No switching between browser and editor. No video scrubbing. No markdown hunting.

Just focused practice. One module at a time.

Real learning happens when you type every command and see every response.

Time & Cost

  • Duration: ~10 hours (4 weeks at 30 min/day)
  • API Cost: ~$10 (actual production API calls)
  • Your time: Priceless (you'll ship AI features)

Commands

# Start/continue course:
ai-terminal-course

# Jump to specific module:
ai-terminal-course 5

# See your progress:
ai-terminal-course

Progress Tracking

Your progress is automatically saved to ~/.ai-terminal-course/progress.json.

Track:

  • Completed modules
  • Time spent
  • API costs
  • Learning streak 🔥

Tech Stack You'll Use

  • APIs: OpenAI GPT-4, Claude 3.5
  • Tools: Helicone, Portkey, Langfuse
  • Patterns: RAG, agents, tool calling, structured outputs
  • Production: Caching, monitoring, cost optimization

Philosophy

  • ❌ No theory lectures
  • ✅ Applied skills only
  • ❌ No copy-paste
  • ✅ Type everything
  • ❌ No fake demos
  • ✅ Real API calls with real costs

You'll understand why things cost what they do, how to optimize, and what production actually looks like.

After This Course

You'll be able to:

  • Ship AI features to production
  • Debug LLM issues
  • Optimize costs and latency
  • Build RAG systems
  • Create AI agents
  • Monitor and improve AI quality

Development

Want to contribute or run locally?

# Clone repo:
git clone https://github.com/josharsh/ai-terminal-course.git
cd ai-terminal-course

# Install:
npm install

# Run:
npm start

License

MIT

Stop reading. Start building.