Artificial Intelligence Course
AI for developers: machine learning basics, large language models, tools and MCP, AI agents, and AI-assisted development with Claude Code.
17 lessons in 5 categories
AI Fundamentals
What Is Artificial Intelligence?
A clear map of what AI actually means -- narrow vs. general AI, how it differs from traditional software, and where Machine Learning, Deep Learning, and Generative AI fit.
Machine Learning
What Machine Learning actually is -- the difference between training and inference, the three types of learning (supervised, unsupervised, reinforcement), and how overfitting and underfitting can make training go wrong.
Deep Learning
How neural networks are structured -- layers, weights, activation functions -- how training works via backpropagation, and a brief look at CNN/RNN/transformer architectures.
Generative AI
How Generative AI differs from discriminative models, how text/image generation actually works, a preview of Large Language Models, and the critical distinction between fluency and correctness.
Large Language Models
How Large Language Models Work
What an LLM mechanically is, pretraining and knowledge cutoff, the difference between base and instruction-tuned models, and a first look at in-context learning and context.
Tokens and Context Windows
What tokens and tokenization are, the hard limit of the context window, what happens when context fills up, and truncation/summarization/retrieval strategies for managing limited context.
Prompting and Prompt Engineering
Prompt structure (system/user/assistant roles), zero-shot and few-shot prompting, effective prompt-writing practices, and an introduction to common techniques like chain-of-thought.
LLM Capabilities and Limitations
What LLMs are genuinely strong at, why hallucination happens, the effect of knowledge cutoff, reasoning limits, and how bias in training data gets reproduced.
Tools & MCP
Tools and Function Calling
The tool use / function calling mechanism that lets an LLM request real code execution: the tool-calling loop, defining a tool, and how tool use differs from an "agent."
Introduction to MCP
What the Model Context Protocol (MCP) is, the N x M integration problem it solves, the host/client/server roles, and the tools/resources/prompts primitives a server can expose.
MCP Architecture
The JSON-RPC 2.0 message format MCP uses, the stdio and Streamable HTTP transports, the initialize/discover/invoke connection lifecycle, and capability negotiation.
Building an MCP Server
Building a real, runnable MCP server and client with the official TypeScript SDK: project setup, defining tools with registerTool, connecting over an in-memory transport, and real output.
AI Agents
What Is an AI Agent?
What defines an AI agent: the observe-decide-act loop, how it differs from a single tool call, and the autonomy spectrum.
Agent Planning and Reasoning Patterns
Concrete patterns for an agent's "decide" step: ReAct, plan-and-execute, reflection, and when a loop should actually stop.
Controlling Agent Behavior
The guardrails that make running an agent loop safe: step limits, human-in-the-loop approval, least privilege, and observability.
Building an AI Agent in TypeScript
Build a real, running agent loop: a deterministic, explicitly simulated decision step genuinely calling the real MCP tools from "Building an MCP Server."
AI Development Tools
AI-Assisted Software Development with Claude Code
Analyze, plan, implement, test, and review a real Spring Boot feature in a real Claude Code terminal session: Plan Mode, the permission model, and bugs actually caught along the way.