Nate's Zero to AI

Nate's Zero to AI

Build your first 5-agent team on Codex with me — AGENTS.md, subagents, and parallel runs explained (no code needed)

Nate's avatar
Nate
Jul 01, 2026
∙ Paid

TL;DR: Once Codex is set up, the real win isn’t one-off help, it’s never re-explaining the same job twice. I take you step by step from a single AGENTS.md file to a custom subagent to Codex running five agents in parallel. And the guide is agent-readable: paste the copy-paste prompt into Codex or any agent and it walks you through the whole build.

The first few runs feel great with Codex: point it at a folder, tell it the job, watch it work. Then you notice you’re retyping the same instructions: the same folder rules, the same “here’s what I want and where to put it,” every single session.

This piece picks up where the Codex setup guide left off, turning that basic working pattern into something reusable.

I’ll take you through it step by step: first we build an AGENTS.md file that gives Codex one standing job, then a custom subagent that owns a single task, and then I push Codex to see how far this goes, asking it to run five agents in parallel and hand back five files at once. Some of what it did genuinely surprised me. I left those moments in.

The example is editorial because it’s easy to watch: Codex plays researcher, takes a rough article idea, and hands back a working brief before anyone writes a word. But the shape underneath is always the same: source in, a couple of ordered passes, finished artifact out. That shape fits client research, weekly reporting, sales notes, or meeting prep just as well.

One more thing worth saying before the list, because it changes how you read the rest: these guides are built to be agent-readable. You’re not meant to read this and then go build alone from memory. There’s a copy-paste prompt you drop into Codex, and it walks you through the setup one rung at a time, pausing after each so you can check the receipts before it goes further. It’s a guide you read and a guide your agent runs.

Here’s what members get:

  • A more complex AGENTS.md file that moves the same work through many roles.

  • The difference between AGENTS.md and Codex custom subagents.

  • One custom subagent for a particular role.

  • An orchestrated team of custom subagents.

  • How to run five parallel subagent workflows.

  • The copy-paste prompt you drop into Codex to run this whole method on your own recurring workflow.

User's avatar

Continue reading this post for free, courtesy of Nate B. Jones.

Or purchase a paid subscription.
© 2026 Nate B. Jones · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture