Coding with AI Agents

AI-assisted coding is everywhere, but most developers are only scratching the surface. This class cuts through the noise and focuses on fundamentals: working with a coding agent end to end, from research through iteration. Every participant brings their own coding agent –- Claude Code, Codex, pi, or whatever they're already running –- and works hands-on with it throughout the session. By the end, participants won't just know how to prompt better; they'll have built reusable skills that compound over time, turning their agent from a one-shot helper into a persistent part of their development workflow.

• The core loop: research → plan → implement → evaluate → iterate
• Skills and MCP connections
• Reviewing output with taste
• Building persistent skills for your agent

• Apply a structured loop when working with a coding agent instead of ad-hoc prompting
• Configure and use skills and MCP connections to extend agent capabilities
• Use sub-agents to parallelize work
• Critically review agent output and iterate until it meets quality standards
• Build at least one reusable skill that compounds value across future projects

• Comfortable writing code in your primary language
• Familiarity with at least one coding agent (Claude Code, Codex, pi, Cursor CLI, Muse Code, etc.)
• Bring your own laptop with your coding agent of choice set up and running