EXTENDED COURSE

Codex

Codex is OpenAI's agentic coding tool — a terminal agent (plus a VS Code/JetBrains extension and a managed cloud) that reads your repo, edits files, runs commands, and iterates on failures until they pass. This hands-on course goes from your first terminal task to a pro workflow: approval modes, project rules in AGENTS.md, choosing models, debugging with git, and the IDE — then deeper into cloud delegation, MCP, headless `codex exec`, and the plans that fit.

1Lessons11step-by-step, ~90 min each

Get started

Work like a pro

Go deep

2Cheat sheetcopy-ready expressions

3Examples3what people built

Internal tools & ops1

Codex Scientist AI-native SME

Course starter

Build a CLI on top of an existing Python library

Codex reads an existing analysis library and generates a Click-based CLI wrapper with subcommands for each public function, help text derived from docstrings, and integration tests — all added to the repo without touching the library itself.

Lab members run analyses from the terminal without writing any Python; the CLI ships as part of the next package release.

Try it yourself

Run `codex`: "Add a Click CLI in cli/main.py that exposes all public functions in src/analysis/pipeline.py as subcommands. Derive help text from existing docstrings. Add tests in tests/cli/."

Research & data tools2

Codex Scientist

Course starter

Refactor a monolithic CSV processing script into modules

A 2 000-line data-cleaning script is split by Codex into four focused modules (ingest, validate, transform, export), with a clean public API and no change to observable behaviour — verified by re-running the original outputs.

The script becomes maintainable and reusable; a colleague adapts the ingest module for a new instrument within an hour.

Try it yourself

Run `codex` in your project: "Refactor scripts/clean_sequencing_data.py into a package with ingest, validate, transform, and export modules. Preserve all existing behaviour. Add docstrings."

Codex Founder Small biz

In the gallery

Explain and fix a failing test suite after a dependency upgrade

After bumping a major dependency version, 14 tests fail. Codex reads the diff, the failing test output, and the dependency changelog, then applies the minimal set of code changes to restore all green — explaining each fix in plain language.

The upgrade unblocks in under an hour with a clear audit trail of why each fix was made.

Try it yourself

Run `codex` with the test output pasted in: "The following 14 tests broke after upgrading pandas to 2.x. Here is the output. Fix the source code (not the tests) to make them pass."