
AI-Native Software Engineering with Coding Agents
Udemy ยท English
About this course
- Language
- English
- Category
- Provider
- Udemy
- Data updated
Coding agents now write a large share of the first draft of software. Most teams adopting them hit the same problems: pull requests too big to review, tests that pass for the wrong reason, agents that ignore conventions, keys in context files, and managers asking whether any of it made delivery faster. This course teaches the engineering discipline that makes coding agents work in a professional team. You will learn to give an agent the context and specification it needs, drive the work with tests it cannot argue with, review its output like a senior engineer, keep it inside safe limits, use it on legacy code, run it in your CI pipeline, and measure whether it improved delivery. You work throughout on OrderHub Brownfield, a realistic codebase with history: a modern Python service and an older Java billing module with a nightly batch and undocumented rules. Every lab has a starting point and a solution, and acceptance tests decide when you are done, whichever tool you use. You will learn to: Write one shared context file (AGENTS md) that Claude Code, Cursor and GitHub Copilot all read, and avoid the precedence trapsTurn tickets into agent-ready specs and plans, and slice work so it can be reviewedDrive features test-first, and catch agents that weaken tests to get a green buildReview agent pull requests risk-first, including performance traps and invented APIsLock down agent permissions and add a CI security gateMap, pin and strangle legacy Java code safely with an agentRun unattended agents in CI with budgets, kill switches and a human merge gateMeasure pull-request size, review time, rework and cost, and run a fair pilotWrite a team agent policy and a 90-day rollout planAbout the tools: the course is tool-neutral at its core. Most live sessions use Claude Code, and the same workflows are shown in Cursor and GitHub Copilot. Tool-specific lectures are dated and updated as the tools change.











