LISTENING TO OUR READERS — HOW I WOULD BEGIN

1. Copy the folder called Listening to Our Readers to a location you use for editorial work. That is the project name. Start with a small set of approved material.
2. Edit the four numbered documents in that folder. A small first trial can put the same information in one document. The separate files are there to make it easy to change things later.
3. In Codex, add or open that folder as a local project and start a chat working there. The exact button labels vary by app version. You do not need to create a software repository to understand this file-based example.
4. The file AGENTS.md holds the standing project instructions. Codex loads instruction files according to its project and directory rules, along with any higher-level instructions. Start with the supplied “Check the project” prompt. Confirm it can identify the intended instruction file and supporting documents. A promise in a chat is not proof that the file loaded. If instructions change, start a fresh chat in the project and check again.
5. For difficult reading across sources, use the strongest reasoning model you can afford. Choose a higher reasoning setting if the tool offers one. Test the result against material you know. A higher price, more thinking, or several agents does not guarantee a better answer. Current model names and controls change; choose from your actual account.
6. Add approved material using the “Add material” prompt. Begin with manual file addition so you can see what is actually being read. Connections to inboxes, forms, or archives require their own permissions, setup, and checks. Nothing in this kit automatically connects them.
7. Use “Prepare a briefing,” then “Check the briefing.” Open the important sources yourself. Ask a different question of the same collection. Bring the briefing into an editorial conversation, prayer, and judgment.
8. Record what you decide, tell readers what happens next when there is something real to report, and repeat. Tweak the instructions and supporting documents when experience shows what needs changing.

Other approved AI tools can use the same editorial rules and documents. Their persistent-instruction and file-link behavior must be checked separately. This kit gives a concrete Codex folder setup; it does not assume that another product automatically reads AGENTS.md.

OFFLINE
All files, prompts, and the worked example are local. Reading them requires no connection. Asking an online AI service for a new briefing requires connectivity and the appropriate account. The saved example is a fallback for demonstration, not proof of a separate live product run.

Official OpenAI documentation checked October 20, 2026:
Custom instructions with AGENTS.md: https://developers.openai.com/codex/guides/agents-md
Desktop app and local projects: https://learn.chatgpt.com/docs/app
Reasoning best practices: https://developers.openai.com/api/docs/guides/reasoning-best-practices
