Show HN: AI Terminal Course – Learn AI Engineering Like Learning Vim (www.npmjs.com)

AI Terminal Course is a terminal-only, hands‑on curriculum that teaches AI engineering by making real API calls instead of showing videos or slides — think vimtutor for LLM production. Delivered as a CLI (npx ai-terminal-course or npm install -g), the course prompts you to supply an OpenAI API key and walks through 20 modules (~30 minutes each, ~10 hours total) for an estimated ~$10 in API costs. Progress is saved locally, exercises are typed live, and everything runs against real models (OpenAI GPT‑4, Anthropic Claude 3.5) and observability tooling. The syllabus focuses squarely on production patterns and pragmatic issues: tokens and LLM calls, streaming, embeddings, RAG, chunking, structured outputs and validation, tool calling and agents (ReAct, planning, reflection), prompt and context engineering, semantic caching, gateways/fallbacks, guardrails/PII detection, and monitoring/cost optimization (caching, model selection, Helicone/Portkey/Langfuse). By forcing live requests and showing real costs, latency, and observability, it aims to close the gap between toy demos and deployable features — useful for engineers who want to ship, debug, and optimize LLM systems rather than just learn theory. The course is open-source (MIT) and includes a capstone to ship a first AI feature.