This is a comprehensive, hands-on program where you will build a fully autonomous AI coding agent from scratch. Over the span of 19 hours, you will engineer a real system capable of reading and understanding an entire codebase, writing and editing files, running shell commands, searching and fetching data from the web, and managing long-running tasks autonomously.
Please note: This is an advanced course and is not suitable for complete beginners.
This curriculum is designed for experienced developers and AI practitioners who already have a solid foundation in software engineering, Python, and large language models. We will work extensively in the terminal and build a production-grade CLI and TUI-based system. Familiarity with APIs, tooling, and modern AI workflows is strongly recommended.
This course is for developers, AI engineers, and system builders who want to move beyond simple chat-based assistants and learn how to design, implement, and scale real agentic AI systems.
What This Course Is About
Modern AI tools can generate code, but real AI agents must do far more. They must plan, execute, adapt, remember, recover from errors, and continue working even when context becomes large or tasks become complex.
In this course, you will build an AI coding agent that:
- Reads and understands your codebase
- Writes, edits, and manages files safely
- Executes shell commands and tools
- Searches and fetches information from the web
- Maintains long-term memory across sessions
- Plans tasks and tracks progress
- Spawns sub-agents when complexity increases
- Detects and breaks out of infinite loops
- Learns from mistakes through feedback loops
- Compacts and prunes context to run indefinitely
- Saves sessions and restores checkpoints
- Allows controlled autonomy and approvals
- Is fully extensible with custom tools and third-party integrations using MCP
This is not a demo project. You are building a real AI system architecture that mirrors how modern autonomous coding agents are designed.
- Architecture and Core Concepts You Will Master: This course takes a deep dive into agentic AI system design, covering both theory and implementation.
- Agentic Loop and Execution Engine: You will design the core agentic loop that allows the AI to plan, execute actions, observe results, and iterate until tasks are complete.
- Tool System and Tool Registry: Learn how to define, register, and expose tools to the AI using schemas, enabling safe and structured tool execution for file operations, shell commands, search, and more.
- Context Management: Implement advanced context management strategies, including compaction and pruning, so the agent remains effective even during long-running tasks.
