Get Started with Codex in 5 Minutes: One Workflow Earned 50K+ on TikTok

If you can only learn one AI tool in 2026, I strongly recommend Codex.

What Can Codex Do?

Many people still think of Codex as just an AI tool for writing code. But in reality, Codex’s task‑execution ability is even more powerful than its coding capability.

Basically, anything you can do on a computer, Codex can do too.

Take myself for example — I run an e‑commerce business, so I built an e‑commerce video editing workflow with Codex:

• Analyze trending videos
• Identify and select素材
• Auto‑dub and add BGM
• Auto‑align subtitles

(Finished CapCut draft output by Codex in one click)

With just this one workflow, my e‑commerce content production efficiency increased 5‑fold, and daily order volume grew by 60%. I haven’t calculated the exact earnings, but at least 30,000‑50,000 RMB.

Besides that, most of my current content output is also powered by Codex. A WeChat article I published a few days ago — on a brand new account — got 1,000 views, which is already a small hit.

And this article — from outline to first draft, including cover and illustrations — was entirely done by Codex. I only made some minor adjustments afterwards. (Not bad, right?)

So, my friends, if you haven’t used Codex yet, you must read this article.

Zero foundation, complete beginner, liberal arts perspective — I’ll explain Codex’s core features one by one. After reading, you’ll know how to use these features and what problems they can solve for you.

Getting to Know the Interface

Open Codex and you’ll see three main areas.

Left sidebar — your conversation list. Basically all your conversations with Codex and your project files are stored here.

At the bottom left, click Settings to check your Codex quota and see how much usage remains. If you run out, you’ll have to wait until the next automatic refresh.

Middle area — the dialog box. You can chat with Codex directly here and ask it to do tasks for you. On the right side of the dialog box, you can switch models and reasoning intensity. Generally, using GPT‑5.5 with medium reasoning can handle most tasks.

At the bottom left of the dialog box, there’s a permission setting with three levels:

• Request approval — Codex asks you before executing any command;
• Review for me — only asks when there’s a risky operation (e.g., modifying document content, deleting files);
• Full access — removes all restrictions, highest automation level but also higher risk. Recommended only after you’re familiar with Codex.

Top right — a button to open the side panel. Whether it’s content generated by Codex or your own project folders, everything is displayed here.

Quick Commands

While chatting with Codex, you mainly use two quick commands: slash (/) and at sign (@).

Slash command: Type a / in the dialog box and a dropdown menu appears. All of Codex’s available functions and skills are listed here.

Want to use a feature? Just type / and it’s there — no need to search around.

At sign command: The @ command also opens a dropdown menu. Unlike the slash command, the at sign mainly invokes plugins, which greatly expand Codex’s capabilities.

For example, the viral trend of creating motion‑effect videos with Codex was done via the HyperFrames by HeyGen plugin.

Another new official plugin is Record & Replay. Once enabled, it records any operation you perform on your computer and packages it into a reusable Skill.

Besides calling plugins, the @ command has another very important use: you can reference your conversations and files, so Codex can quickly locate the content you want to modify.

For instance, with this article I’m writing, I need Codex to review it and suggest improvements. I just type @ + the file keyword to find the article quickly.

Project Management

After reading the first two sections, you should have a basic idea of how to use Codex. But if you want it to solve real problems, you’ll definitely need the Project feature.

Simply put, Codex creates a corresponding folder on your computer. All files it generates are automatically saved into that folder.

How to create one? Click Enter Project Mode below the dialog box. You can create a blank folder or directly use any existing folder on your computer.

This feature is really practical. Before, when using the chat mode, I had to manually download or copy generated files and often forgot where I saved them. With Projects, all files are in one place — open the folder and you see everything.

Moreover, in Project mode, you can create multiple conversations within the same project. For example, ask Codex to organize data in one conversation and write copy in another — both tasks run simultaneously, and generated files are all stored in the same project folder.

I use Codex mostly in Project mode, unless I have a quick small question — then I’ll start a new conversation for a casual chat.

Task Execution

When we used the web version of ChatGPT before, there were two awkward problems:

1. You could only have one dialog box open at a time — you had to wait for one task to finish before starting the next.
2. If you saw the AI was going in the wrong direction, you had to stop and re‑enter the instructions.

But Codex provides good solutions for both.

First, Codex can handle multiple tasks simultaneously.

Once you assign a task, you don’t have to wait for it to finish. You can start a new conversation and assign another task right away. As long as the tasks don’t modify the same document, they won’t interfere.

When the circle next to a conversation turns into a small dot, that means the task is done — you can go check it.

Within the same conversation, you don’t have to wait for Codex to reply. You can send new instructions at any time.

By default, new instructions are executed in order.

You ask it to do Task A. While it’s running, you can directly send Task B. Codex will finish Task A first, then automatically proceed to Task B, then C, then D…

If you notice Codex is going in the wrong direction, you can jump the queue. Click the guidance button, and the new task starts immediately without waiting for the previous ones to finish.

Custom Instructions

Open Settings, find Personalization on the left, and you’ll see a field for custom instructions.

Here, you can tell Codex in advance what to pay attention to when communicating with you.

My own setup is roughly: ask Codex to communicate in plain language, avoid technical jargon, and check its own work first before coming to me with questions.

You can also write according to your habits — for example, if you want concise replies, or want it to always provide a checklist.

Creating Skills

The purpose of a Skill is to freeze repetitive, complex processes into a reusable template. Next time you perform the same task, you can just invoke the Skill and execute it with one click.

For example, the e‑commerce video editing workflow I mentioned at the beginning is a Skill. When I use it, I just type a slash command (type / to see all Skills), and Codex follows the pre‑designed process. A few minutes later, I can see the edited video in my CapCut drafts.

How to create one? First, let Codex run through the entire workflow. If you’re satisfied with the result, tell it to turn this workflow into a Skill. Codex will automatically save it, and you can invoke it quickly next time.

You can also use Skills created by others. Just send the Skill’s address to Codex and ask it to install the Skill.

For instance, Kazike’s AI Writing Skill, AI Daily Brief Skill, and Guizang’s PPT Skill are all excellent Skills.

Automated Tasks

Once you’ve mastered the sections above, you can try letting Codex execute automated tasks.

For example, I receive a daily briefing prepared by Codex — it covers noteworthy AI tools, news, knowledge management, and personal growth content from across the web, all delivered to me first thing.

I also have an automated Topic Brief task. Codex reads the daily briefing every day, identifies potential writing topics, and converts them into requirement briefs stored in my knowledge base. (The article you’re reading now was also a topic suggested by Codex.)

How do you set up an automated task?

It’s very simple. After you’ve successfully run a complex task, just tell Codex: “Execute this task automatically every day at 9 AM (or every week).”

Codex will create the automation for you.

If you want to set up a truly useful automated task, it might be a bit more complex.

Take the AI Daily Brief automation — it involves when to invoke a Skill, what tool to use to gather information, where to save the collected information, and how to notify me after saving.

You need to describe every step in the workflow very clearly, even turn each step into a Skill, then have Codex automatically invoke it.

This feature is especially suitable for tasks that need to be done regularly, like organizing files, generating weekly reports, or monitoring content updates at fixed intervals.

Desktop Pet

The desktop pet feature is not only fun but also quite useful.

Simply go to Settings > Pets and turn it on to select your favorite pet.

After waking up the pet, it floats on your desktop.

It shows Codex’s working status in real time — moves around when executing tasks, and stops when the task is done.

This means when you ask Codex to do something, you can work on other things. With just a glance at the desktop, if the pet has stopped moving, you know the task is finished. No need to switch back and check.

There are 8 official pets to choose from, and you can also create your own. Click Create Your Own Pet and it will jump to the dialog box.

Then just send your reference image to Codex — nothing else needed. After a few minutes, you’ll have a beautiful custom pet avatar.

Final Words

I’ve covered the main features of Codex.

There are many other features — like browser & computer control, hooks, work trees… If you’re interested, I’ll write an advanced guide later.

But even just reading this article gives you a pretty solid understanding of Codex.

The rest is just playing with it yourself.

That’s all. Hope this helps.

Best regards. 🍀