Most of the content out there is still a rundown of features. It tells you what Codex can do, but not how to combine those features into a useful system, which use cases it’s actually good for, or how to set it up so it becomes part of the way you work.

That’s pretty much the intention of this video: to give you an overview of what Codex can do and the operating system you need to build in order to make it work well. It mirrors how we use Codex inside our company and how we’ve built an operating system underneath it, with Codex features sitting on top.

The important part is that the operating system isn’t tied to Codex. If we decide to experiment with another harness or move to a different system later, the underlying structure stays intact. We can simply plug in a new tool without rebuilding everything from scratch.

It’s extremely beginner friendly, but it still covers nearly every major Codex feature and shows how to use each one in a real system. We’ll cover how to build a strong knowledge layer, work with plugins and skills, create personal evals, and use Codex Threads, slides, and other features. Here’s everything we cover in a 17 min video (phew, that’s a lot now that I see it).

Hope you enjoy the video. If you do enjoy this style of short but knowledge-dense videos, do check out the previous two videos on my channel as well:

• AI Engineering Explained in 17 mins | The Ultimate Beginner’s Guide

• How to Use AI Agents Better

I’ve always been a little averse to roadmaps. They look clean, but they often make learning feel more linear than it really is. Learn this, then this, then this, and somehow you’re supposed to know how to build.

But humans don’t really learn that way. We learn by connecting ideas. We remember through stories, patterns, examples, and relationships between concepts. That’s why mind maps often work better, especially for messy fields like AI engineering.

What I wanted to create was more of a mind map, basically a connected set of ideas you can keep coming back to, even as new concepts come in. Something that builds your foundation instead of confusing you further.

Here’s my attempt to explain what it takes to become a good AI engineer in the most beginner friendly way possible.

And doing all of this through a connected mind-map, over scattered ideas. The goal is to help you build a strong foundation that can take you anywhere, without memorizing endless roadmaps and disconnected concepts.

Here’s a snippet of the topics covered, do watch the video to go deeper (it’s only 17 mins long and super value packed!)

Do check it out, and send it to anyone who’s feeling overwhelmed and struggling to connect the ideas together.

Starting this week, we’ll be releasing new videos every week, across different levels and for different kinds of learners. There’s a lot of tool FOMO on YouTube right now. Every week, there’s a new agent, a new coding tool, a new workflow, and a new “must use this” claim. We want to bring systems thinking back into how people learn AI tools. Less chasing every new tool. More understanding what actually transfers across tools. So if that’s your vibe, consider subscribing. We’re making high quality, beginner friendly, engaging videos that help you understand AI tools from first principles.

Every week there is a new AI agent to learn. Claude Code, Codex, Cowork, OpenClaw, the next one shipping while you read this.

Most people respond by chasing each release, memorizing commands, and starting over when the tool changes. They stay busy and never compound.

I made my first YouTube video about the way out of that loop and it’s super beginner friendly.

Here is the idea: Every agent on the market runs on the same structure. A language model in the middle, four components wrapped around it, and one decision that lives with you.

Learn those once and you stop relearning tools. The next product that lands does not catch you off guard.

Video linked above. Tell us what resonates!

-Aish