AI as a coding assistant

A coding assistant is the same kind of LLM you've been calling all along, pointed at code and wired into your terminal. You describe what you want in plain language and it produces code, or reads code you already have and tells you about it. Nothing new about the model - what's new is that it sits next to your files and can act on them.

What you ask it to do

Almost every request to a coding assistant is one of four kinds:

• Write. Turn a description into working code. "Write a bash script that lists the five biggest files in a folder."
• Explain. Read code or an error and say what it does or why it broke. "What does this awk line do?" or "Explain this stack trace."
• Fix. Point it at a bug and have it propose the correction. "This script fails when the folder is empty - fix it."
• Refactor. Keep the behavior, improve the shape. "Rename these variables to something clearer" or "split this function in two."

These are the same language tasks LLMs are good at from the very first topic, just aimed at source code instead of prose.

It works where your code lives

A chat window makes you copy code out and paste answers back. A terminal assistant skips that: it runs in the same folder as your project, so it can read the files that are already there and write its answer straight to disk.

This VM has two of these ready to go: the claude CLI and the codex CLI, both preinstalled. The next node uses claude.

This is only the surface

Coding assistants go much deeper than a few one-line asks - working across a whole codebase, running commands, following your project's conventions. That's a subject of its own. A separate claude-code-fundamentals skill covers it properly. Here you just get the basics: enough to use claude from the terminal for everyday coding help.

Your lab setup

This VM comes preconfigured with your AI keys - just type claude to get started. Your key and URL are already set as $OPENAI_BASE_URL and $OPENAI_API_KEY.