Coding Agents & AI-Assisted Development
Understand how coding agents work and use them well: the agent loop, tools, edit formats, context, instruction files, test-driven workflows, parallel agents, verification, evaluation, cost and team practice, with a working mini agent you can run.
Syllabus
What Coding Agents Are
- From Autocomplete to Agents
- Anatomy of a Coding Agent
- The Agent Loop in Action
- Where Agents Run: Editor, Terminal, Cloud and CI
The Tools That Make It Work
- Reading and Searching a Codebase
- Editing Files Reliably
- Running Commands and Tests, and Reading the Results
- Git as the Safety Net and the Handoff
Context: What the Agent Knows
- The Context Window as a Budget
- Instruction Files: Teaching the Agent Your Project
- Long Tasks: Compaction, Notes and Fresh Starts
- Sub-Agents and Context Isolation
Working With a Coding Agent
- Writing a Task the Agent Can Succeed At
- Test-Driven Loops: Tests as the Target
- Small Steps, Plans and Review Checkpoints
- Parallel Agents With Git Worktrees
Keeping Agents Reliable
- How Coding Agents Fail
- Detecting Stuck and Runaway Agents
- Patch Gates and Secret Scans
- Security Basics: Permissions, Injection and Least Privilege
Measuring Agents: Quality, Cost and Value
- Benchmarks and Their Limits
- Success Rates, pass@k and Variance
- What a Run Costs: Context Growth and Caching
- Measuring Productivity and Quality Honestly
Using Agents Well in Teams
- Rolling Out Agents to a Team
- Ownership, Licensing and Accountability
- Learning and Career: Staying the Engineer