OpenAI Publishes 'Working with Codex' — A Step-by-Step Guide to AI That Actually Does the Work

OpenAI just released "Working with Codex" on their Academy platform, a comprehensive guide teaching users how to go from zero to productive with Codex — their AI agent that doesn't just chat, but actually completes tasks.

The guide covers the full workflow: setting up a workspace, creating projects with isolated contexts, opening threads to delegate tasks, managing files directly within Codex, and getting finished deliverables back — documents, dashboards, code, and more.

What makes Codex different from a typical chatbot is that it operates as a true agent. Instead of giving you text answers to copy-paste elsewhere, you describe what you need and Codex executes it end-to-end. Upload a spreadsheet and ask for a summary chart — it builds the chart. Describe an API endpoint — it writes, tests, and returns the code.

The timing is significant. OpenAI recently expanded Codex with background execution on macOS and Windows, built-in browsing, image generation, memory, and plugin support. Combined with the new Workspace Agents feature — which lets teams create reusable automated workflows — Codex is positioning itself as the operating system for AI-powered work.

This Academy guide lowers the barrier to entry considerably. Rather than figuring things out through trial and error, users now have a structured path from setup to real output. For teams evaluating AI productivity tools, this is the most practical onboarding resource OpenAI has published to date.

The era of AI that only answers questions is ending. The era of AI that does the work has begun.