Engineering Craft › AI-Assisted Development

AI Coding Assistant

Tools like Copilot, Cursor or Claude Code that suggest and write code.

Also known as: AI assistant, Copilot, Cursor, Claude Code, coding assistant

An AI coding assistant is a tool that uses a large language model to help you write code. It can complete the line you’re typing, answer questions about a codebase, explain errors, write tests, and, in agent mode, make multi-file changes and run commands.

Forms they take:

• Inline completion in the editor, suggesting the next lines.
• Chat beside your code, for questions and explanations.
• Agents that read files, edit several of them, run tests and iterate on a task (coding agents).
• Command-line assistants that work in your terminal.

Examples include GitHub Copilot, Cursor and Claude Code. The products change quickly.

Where they help

• Boilerplate, repetitive edits and scaffolding.
• Explaining unfamiliar code or an error message.
• Drafting tests and documentation.
• Translating between languages or formats.
• A quick first version to react to.
• Working through ideas as a rubber duck that talks back.

Where to be careful

• They can be confidently wrong: invented functions, outdated APIs (hallucination).
• They don’t know your system unless you tell them. Context and conventions matter (agent instructions).
• Security and privacy: don’t paste secrets or private data into tools you haven’t been approved to use. Check your company’s policy.
• Licensing and ownership questions may apply to generated code.
• It’s easy to stop understanding what you ship (learning while using AI).

Using one well

• Be specific about what you want, with examples and constraints (prompting for code).
• Work in small steps, and review each one.
• Read and test everything it produces, as you would a stranger’s pull request (reviewing AI code).
• Ask it to explain, then verify against the official docs.
• Stay the author. You’re responsible for what you commit.