AI coding tools

This course encourages you to use AI tools to support your learning — as long as you take responsibility for what you submit and learn from the output. See the syllabus for the full policy.

Below are some AI coding tools worth knowing in 2026. They overlap significantly in what they can do; pick one or two and get comfortable with them rather than spreading thin across all of them.

Important: AI assistants are useful, but you are responsible for understanding every line of code you submit — especially because the oral exam will test your understanding, not the AI’s.

Note on access: BI students get free access to Google’s Gemini models through their BI student email, but this does not include the coding tools such as Gemini CLI or Antigravity. Most tools listed below offer a free tier, but the usage limits are tight. A paid subscription (typically around $20/month) gives a much better sense of what these tools can actually do. One exception worth applying for: GitHub Copilot Pro is free for verified students via the GitHub Student Developer Pack — see the Copilot section below.

Claude Code

Claude Code is Anthropic’s official command-line agent for software engineering, built around the Claude model family. You run it inside your project folder and it can read, write, and execute code on your machine.

  • Terminal-based
  • Strong at multi-file refactors, debugging, and explaining unfamiliar codebases
  • Available in editors too via the VS Code and JetBrains extensions
  • Docs: docs.claude.com/en/docs/claude-code

Cursor

Cursor is an AI-integrated code editor built on top of VS Code. It keeps the editing experience similar to VS Code but adds three core AI features:

  • Tab autocomplete — aggressive multi-line completions as you type
  • Chat — ask questions about your code in a sidebar
  • Agent mode — give it a task and let it edit files across the project

A good choice if you want AI assistance without leaving a familiar editor. Free tier available; paid tier for heavier use.

Codex

Codex is OpenAI’s open-source coding agent. It comes in two flavours:

  • Codex CLI — a terminal agent similar in spirit to Claude Code
  • Codex IDE extension — VS Code integration

Powered by OpenAI’s GPT-family models. The CLI is open-source on GitHub and installable via npm. Useful if you already use ChatGPT and want a more code-focused agent.

Gemini CLI

Gemini CLI is Google’s open-source terminal agent, powered by the Gemini model family.

  • Free tier with generous usage limits for personal Google accounts
  • Built-in tools for file edits, shell commands, and web search
  • Installable via npm install -g @google/gemini-cli

A reasonable choice if you want to compare different model families against the same task — running Claude Code, Codex, and Gemini CLI side-by-side on the same problem is a good way to develop intuition for their strengths.

GitHub Copilot

GitHub Copilot is GitHub’s AI pair programmer, integrated directly into most major editors. Like Cursor, it lives inside your editor; like the others, it can also act agentically across multiple files.

  • Tab autocomplete in VS Code, JetBrains IDEs, Neovim, Visual Studio, and more
  • Chat for questions, explanations, and refactors
  • Agent mode for multi-file edits and longer tasks
  • Lets you pick between model families (Claude, GPT, Gemini) on a per-task basis

Free for students. Verified students get Copilot Pro for free through the GitHub Student Developer Pack. Sign up with your BI student email and student ID — this is the best value in this list if you already work in VS Code or a JetBrains IDE.

How to choose

There is no single “best” tool — they all do roughly the same things and improve every few months. Suggestions:

  • Already in VS Code or JetBrains? GitHub Copilot (free for students) or Cursor.
  • Live in the terminal? Claude Code, Codex CLI, or Gemini CLI.
  • Curious about model differences? Try two of them on the same problem.

What matters more than the tool choice: develop the habit of reading what the AI produces, asking it to explain anything you don’t understand, and verifying its output against documentation or tests.

MST0052 Predictive Modelling with Machine Learning · Fall 2026 · BI Norwegian Business School