The Complete Codex Masterclass: Zero to Hero

Become an Agentic Engineer by mastering the OpenAI Codex agent. From zero experience to advanced workflows!

Created byShawn Esquivel

Last updated 9/2026

English

What you'll learn

Plan, build, review, and deploy a web app with Codex.
Delegate focused project tasks to subagents and review their results.
Turn a recorded workflow into a reusable skill with Record and Replay.
Plan and assemble a launch video using Codex and supporting tools.

Course content

4 sections • 32 lectures • 1h 55m total length

Build an AI Jobs Dashboard: Course Project Overview1:55
Plan the MVP and build an AI jobs dashboard to surface jobs, compare pay, and identify recurring skills, then set up the codex project and deploy for user traction.
Choose a Tech Stack That Works Well with AI Agents5:37
Set Up Your Codex Project and Install UI Skills3:34
Prompt for next section0:38
Plan and Scope Your MVP Before You Build5:36
Prompt for this section
# GOAL
We're building an AI Job board.

## Theme
Dark, modern SaaS

## Layout
### Home Page
CTA / hero

### Carousel
just a carousel that smoothly scrolls horizontally, and pauses whenever the user puts their mouse over it. shows all theo jobs randomly.

### Job Table
- Job Board. A table with every job.
-> Link
-> Title
-> Description
-> Estimated Pay

Interactions
1. Click should expand the job modal. Esc should get out of it.
2. Filtering by keyword, location, type of job.

# MVP
- Start with 5 from each data source, using Web/Browser.
- Anthropic, OpenAI, Cursor
- Store everything locally for now
- DONT add a way for people to apply, or upload resume
- Don't do auth
- Don't do dark mode

### TECH STACK
Next.js
TypeScript
Tailwind CSS
shadcn/ui
Vercel

> Install all dependencies as needed.

### Add these skills before starting too, and use them.
- shadcn skills
- make-interfaces-feel-better skill
Build and Refine the AI Jobs Dashboard MVP6:00
Save Your Codex Project to a Private GitHub Repository3:45
Deployment Option 1: Deploy with Vercel CLI4:32
Deployment Option 2: ChatGPT Sites3:00
Review Code with Agents and Codex Cloud10:12
Review before release
Schedule Nightly Code Reviews and Wrap Up the Project4:18

Why Use Subagents? Context, Focus, and Parallel Work5:59
See how a broad app review can be divided into focused tasks. The lesson diagram uses this illustrative request: “Review my AI jobs dashboard.” Task A checks keyboard accessibility; task B finds important missing tests; task C checks for slow behavior and clicks through the app. Use the same structure for your own project, with a clear result requested from each subagent.
Why use subagents?
Your First Subagent Delegation in Codex2:08
Explore how to delegate tasks with subagents in Codex, spawn agents, and run parallel tool calls across GPT 6.0, while managing tokens and waiting for results.
Set Default Models and Reasoning for Subagents2:32
Choose subagent settings
Copy this prompt for the next video0:49
Prompt for this video

Delegate this work out to subagents where possible using GPT 5.6 Sol High
As the sub-agents finish their jobs, you can stitch together all of the pieces together until everything looks good.
Also, we'll just focus on Anthropic, OpenAI, and Cursor for now, and we just want to get everything running end to end before we make everything super production ready.
Copy these prompts: Plan a feature and add Neon and Drizzle0:36
Plan a Feature and Add Neon + Drizzle to an App7:54
Lecture: Plan a Feature and Add Neon + Drizzle to an App

PROMPT 1 — Plan the daily jobs feature
Shown around 00:30–01:00. Exact text recovered from the original Codex chat:

I want this to check the AI labs every single day, and update for major job postings. What would that take? We just want the "AI jobs" specifically; it should be technical or semi-technical.

Help me plan this feature out.

PROMPT 2 — Change the implementation to Neon and Drizzle
Shown around 02:50–03:15. Exact text recovered from the original Codex chat:

Let's use Neon CLI to start a database, and use Drizzle as the ORM for the jobs.

Context: Prompt 2 follows the planning discussion in the same chat. Around 03:49 the instructor separately suggests asking Codex to install the Neon CLI and help with login; that spoken example is not one of the two submitted prompts above.
Build a Feature in Parallel with Subagents5:37
Watch a feature build divided among focused subagents. The copyable prompt shown in this lesson is attached as a text resource. It assumes you already have a feature plan in the Codex chat; add your own plan before reusing it. The original example asks for GPT 5.6 Sol High, which was the model named in the recording—choose an available model for your current setup.
Steer, Queue, Fork, and Ask Side Questions in Codex3:10
Steer, queue, and ask a side question
Copy these prompts for the next video: Custom Agents0:52
Create Custom Agents for Code Mapping and UI Debugging5:31
Lecture: Create Custom Agents for Code Mapping and UI Debugging
Source video: Subagents 7 - Part 1 - Custom Agent Review.mp4
Official Codex reference (see "Example 2: Frontend integration debugging"): https://learn.chatgpt.com/docs/agent-configuration/subagents

PROMPT 1 — Create the custom agents
Shown around 01:40–02:00. Exact text recovered from the original Codex chat:

Create the Example 2: Frontend integration debugging custom agents exactly as per OpenAI specs

Then use them that agent setup to figure out why some of the average pays are not showing.

Context: The instructor pasted the Codex Subagents documentation into this chat before sending this prompt. The prompt refers to Example 2 in that documentation. The original phrasing above is preserved, including its typo.

PROMPT 2 — Delegate the Market Signals debugging task
Shown around 03:20–03:45. Exact text recovered from the original Codex chat:

Investigate why the Market Signals → Average annual pay table shows “Not available” and 0% coverage for Cursor, xAI, Mistral AI, and Composio, while other companies display averages.

Workflow:

- spawn browser debugger and code mapper to reproduce the issue / map out the path
- use ui fixer to make the fix

Create a pull request for this, don't commit to main

Context: The instructor attached a screenshot of the broken table to this debugging request. Students should attach a screenshot of their own issue when adapting the example.
Give custom agents distinct jobs
Subagents Wrap-Up: Models, Reasoning, and Custom Agents1:23
Wrap up on subagents shows how to trigger subagents, assign custom models and reasoning, and use code mapping, browser debugging, and UI fixer to automate tasks and ship code faster.

Copy prompt for next video: Develop the Launch Video Concept and Story0:47
Develop the Launch Video Concept and Storyboard6:21
Turn a launch-video goal into a research brief, a short founder interview, and a storyboard. Download the copyable planning prompt shown in this lesson. It includes a Higgsfield API option; using that service may incur charges, so adapt the resource list to tools you actually have before running it.

THE PROMPT:
GOAL:
We want to create a founder launch video to promote our app. 30-60 seconds long. Less is more, focus on creating on emotion and storytelling.

RESOURCES:

• Me, the founder. I can record a 10-30second, talking head video.
• Actual screen recordings of the app at play when I'm talking over those features using Remotion CLI.
• Higgsfield API. You can plan out a single animation that is inserted strategically somewhere throughout the video. Beginning, middle, or end, doesn't matter but it should tastefully add to the video.
• FFMPEG and other video editing tools - you can do the final assembly, adding graphics, text on screen, etc. to make a polished video.

STEPS

1. RESEARCH: Before we create the video, research what makes a good launch video in terms of structure, copy, sound, visual direction etc.
2. INTERVIEW: Given our constraints and tools available - ask me 3-5 questions to hone in on the direction.
3. STORYBOARD: After we lock in a direction; let's map out the scenes, copy, and animations on a simple HTML page so I can approve the direction - just like a film director would story board it.
Record the Founder Segment for Your Launch Video1:32
Plan a Product Launch Video with Codex, Remotion, and Higgsfield2:41
Assemble and Refine the Launch Video with Codex4:35
Publish Your Launch Video Across Social Platforms1:45
Grab the prompt for the next video: Motion Design Showreel0:07
Codex for Motion-Design Showreel1:08
THE PROMPT:

make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out but keep it on brand

Requirements

A computer and access to Codex. Follow-along deployment and video exercises may require accounts for GitHub, Vercel, Canva, or Higgsfield.

Description

Become an Agentic Engineer by mastering OpenAI's Codex agent. From zero experience to advanced workflows!

I built this course around a real SaaS idea - an AI Jobs Directory. With Codex, we go from idea, to first prototype, to even generating a launch video to get our first users.

You’ll see my iterative process with Codex, from how I use Skills sharpen its outputs to using Subagents to speed up my workflows. Follow along with the course project, or extract my Codex workflows for your own wild ideas.

In the course, I’ll show you how to:

Choose a useful first version of your app so you can start building without planning every feature.
Give Codex clear instructions, check what it builds, and improve the result.
Save your project on GitHub so you can track changes and keep working on it.
Deploy with Vercel so you have a live link to share.
Use ChatGPT Sites to deploy your site to users.
Review your app with Codex and agents to find problems you might have missed.
Schedule regular reviews so you can keep improving the app after launch.
Turn a recorded workflow into a Codex skill you can use again.
Use subagents to split a larger job into smaller, focused tasks.
Make a short launch video that shows people what your app does.

Throughout the course, I'll drop core Codex concepts, so that you can get a hands on feel for what it takes to build an app from scratch, then maintain that app over time with Codex.

Pre-requisites

No professional programming experience required, all you need is a ChatGPT subscription!

About Your Instructor

Shawn is also an OpenAI Codex Ambassador, throwing workshops and events for the Codex community.

Shawn Esquivel is a bestselling Udemy instructor having taught over 3000 students how to use Agentic Coding tools like Codex, Cursor, and LangChain.

Who this course is for:

Beginners exploring AI-assisted app building, and creators who want to connect development with repeatable workflows and launch content.

The Complete Codex Masterclass: Zero to Hero

What you'll learn

Explore related topics

Course content

Build a Web App with Codex11 lectures • 49min

Record and Replay: Reusable Skills2 lectures • 11min

Work Faster with Subagents11 lectures • 37min

Make a Launch Video with Codex8 lectures • 19min

Requirements

Description

Who this course is for: