Teach Your AI to Edit Like You (codex/claude Code System)

In this guide, you will learn how to teach Codex or Claude your editing style by turning finished edits into reusable writing rules. The real move is the loop: draft, snapshot, edit, compare, improve.

Instead of asking an AI to "remember my style," you will give it a small file system that makes your style visible. It drafts an email, freezes the first version, waits for your final edit, then compares the before and after to improve the rulebook for next time.

• Creators and marketers who draft repeatable email, newsletter, or sales copy
• Operators who want AI to improve from real work instead of one-off prompt tweaks
• Anyone building agent workflows who needs a simple example of skills, files, statuses, and automation working together

You will set up a copywriting project structure that gives your AI a place to draft, save final edits, and learn from the difference. The folder keeps your source material, draft copy, finished copy, editorial rules, reusable skills, and automation instructions separate so the workflow does not turn into one messy prompt thread.

By the end, you will have a small working system: one skill drafts copy from your rules, another compares the draft against your final edit, and an automation can periodically update the rules from new examples.

• Codex, Claude, or another agent that can read and edit project files
• A project folder for the email system
• A few opinions about your writing style, or examples of copy you like
• One email assignment you can use as the first test draft

This same pattern works outside marketing emails. Use it for scripts, newsletters, landing page copy, sales follow-ups, social posts, or any format where you repeatedly edit AI drafts.

The reusable loop is:

1. Keep a rules file.
2. Create an AI draft.
3. Freeze the first version.
4. Edit the final yourself.
5. Compare before and after.
6. Update the rules.
7. Automate the next review.

After a few weeks, look for friction. If you keep forgetting to update the status, make the status easier. If the folder structure feels too heavy, simplify it. If the rulebook gets noisy, ask the agent to deduplicate and merge similar rules.

The goal is not to build a perfect editorial operations system on day one. The goal is to make every finished edit teach your AI how to write one percent closer to your style next time.