Coding with AI Assistants: Copilot, Cursor and Claude Code
Ship faster with AI pair programmers — and keep the code correct.
Taught by Anisha Rai · Senior Software Engineer
What you'll learn
- Choose between inline completion, chat and agent modes in Copilot, Cursor and Claude Code for a given task
- Write task briefs that give an agent the goal, constraints and acceptance criteria it needs
- Set up rules files and repository conventions that keep AI output consistent with your codebase
- Review AI-generated diffs for hallucinated APIs, scope creep and silent behaviour changes
- Run a test-first loop where the agent implements against failing tests you control
- Apply security practices for secrets, prompt injection, dependencies and permission modes
About this course
AI coding assistants went from autocomplete to autonomous agents in about two years, and most developers are still using them like a faster search engine. This intermediate course is for working software engineers who want to get the real productivity gains of GitHub Copilot, Cursor and Claude Code without shipping bugs, security holes or code nobody understands. You will learn how each tool fits a workflow, how to frame tasks so an agent does the right thing first time, how to manage context with rules files and repository conventions, how to review AI-generated diffs with the rigour of a senior engineer, how to use tests as guardrails in an agentic loop, and how to handle secrets, prompt injection and team policy. Every lesson uses real commands and configuration files. You should be comfortable with Git, a modern language and writing unit tests.
Course content
6 lessons- 1Completion, Chat and Agents: The Tools and Where They FitHow Copilot, Cursor and Claude Code differ, what an agent actually does, and how to pick a mode per task.Preview
- 2Framing Tasks So the Agent Gets It Right First TimeWrite briefs with goal, constraints, acceptance criteria and a plan step before any code is changed.
- 3Context Management: Rules Files, Conventions and MemoryGive assistants standing instructions with CLAUDE.md, Cursor rules and Copilot instructions, and keep context lean.
- 4Reviewing AI Diffs Like a Senior EngineerA review checklist for hallucinated APIs, scope creep, silent behaviour changes and code you cannot explain.
- 5Tests as Guardrails: A Test-First Loop With an AgentWrite the failing test yourself, let the agent implement, and make passing tests the definition of done.
- 6Security, Secrets and Team Policy for AI CodingPrompt injection, secrets in context, hallucinated dependencies, permission modes and a team policy that sticks.