Map your ecosystem

3.07[BONUS]

What the app already knows about you

Your AI already knows you, and it can hand that over.

Months of chats, and any projects you have set up, have taught ChatGPT or Claude who you are, what you do, the people you mention and how you like things done, and that picture is context for the map and raw material for the brief.

In the app

Who you are
What you do
The people you mention
How you like things done
What you are working on

Held inside the app, and you have never read it.

In your folder

ecosystem/

about-me.md
my-work.md
key-people.md
how-i-like-things-done.md
my-work-now.md

Written down, yours, and read at the start of every chat.

The prompt below gets the picture out.

Run it in a fresh chat in whichever app you have talked to longest, with memory switched on, as many times as you like, because each pass surfaces things the last one missed.

Get the picture out

You have a picture of me built up from everything we have ever talked about, and I have never seen it written down. I want it out, as documents I can keep and use anywhere. First, empty it out. List everything you know or remember about me, word for word where it is stored, grouped into who I am, my work, the people around me, how I like things done, and what I am working on now. Under how I like things done, include every instruction I have given you about how to answer and every correction I have made to you, in my own words. Include what is in any projects I have set up with you, the instructions and the documents in them, each under the area of my life or work it belongs to. Mark each line as something I told you or something you worked out, be exhaustive rather than tidy, and do not soften or leave things out to be polite. Then interview me. Take the groups one at a time and ask three or four questions per round, to correct what is wrong and fill what is missing. Where you are not sure of something, ask rather than assert, and keep going until each group could brief a capable new assistant on its own. Then write the documents. One per group, in markdown, each complete on its own, named about-me, my-work, key-people, how-i-like-things-done and my-work-now, and one more per project, named for the project, holding what that project knew. Plain facts in the present tense, nothing flattering, and nothing I told you to drop. At the top of each one, say in one line which area of my life or work it belongs to, and put each document in its own code block so I can copy it cleanly. Finish by telling me whether that is everything you hold about me or whether anything is left, giving me anything that is, and naming anything in the documents I should think twice about keeping in a file at all.

Note: Read every document before you keep it, because these files are going to tell your agent how to behave, and a wrong line in them becomes a wrong instruction it follows.

Then bring the documents home.

Open Claude Code in your ecosystem folder, paste this in and paste the documents after it, and it saves each one into to-file.

Bring them home

I have documents about me that another AI wrote from everything I have ever talked to it about, and I am going to paste them in. We are inside my ecosystem folder. Save each document as its own markdown file, named as at the top of it, in the to-file folder, and nowhere else yet. When they are all in, read them back and tell me what you now know about me and about each area. Point out anything that contradicts something else in the folder or looks wrong, because these files will shape how you behave, and I would rather correct them now than find out later.

Then paste in the filing prompt from 3.05, and they go to the right areas with everything else, so the brief interview in stage four finds them there.