Understanding AI

You became your AI's assistant

The fix is boring and it changes everything. One folder structure you actually work out of.

What you'll walk away with

A prompt that interviews you and writes the whole structure, every file in full
The shape of the tree, one top folder of standing facts and a branch for each part of your work
How to work out of it in the Claude desktop app, Claude Code, or a Project
The one habit that keeps it alive, so you never re-explain the same thing twice

About twenty minutes, most of it answering questions.

The Idea

Right now you paste the same background into every chat, carry answers between apps, and do the remembering for a thing that is meant to remember for you. A folder structure flips it. Everything an AI would have asked you for is already written down and filed where it belongs, so you open the folder, ask, and it goes and finds what it needs.

The shape of it

my work/ context.md who I am, what I run, how I like things done client work/ context.md the clients, how the work runs, what good looks like personal work/ context.md the house, the admin, the money, the diary content/ context.md what I make, who it is for, where it goes writing style/ the writing folder, one branch of its own

One folder at the top with the facts any AI should know about you. A branch for each part of your life, each holding what you'd hand a new starter hired for that part. Yours will have different branches, and that is the point.

Build Yours Now

One paste. It interviews you, picks two or three branches to start with, and writes every file for you. About twenty minutes, most of it answering questions.

Paste this into your AI

I want you to help me build a folder structure that means I never have to re-explain myself to an AI again. It'll live on my computer, and I'll open it whenever I want your help, so you already know who I am and how my work runs. Interview me one question at a time, in plain English. Don't write anything until you've asked enough. First, find my branches. Ask about the parts of my life and work where an assistant would genuinely save me time, then suggest two or three folders to start with. Not more, because I have to actually fill them in. Then the top folder. Ask me for the standing facts any AI should know before helping me with anything. Who I am, what I do, what I run, how I like things done. Then each branch in turn. Interview me the way you'd brief a new starter hired to help with that part of my life. How it works, who's involved, how things get done, and what someone guessing would get wrong. When you have enough, write the whole thing. Show me the folder tree, then every file in it written out in full. Keep each file short enough to read in a minute, and leave nothing for me to fill in. Finish with two things. Tell me which one or two of my answers were thin, so I know where the first gaps will show. And remind me of the rule that keeps this alive, every time I catch myself re-explaining something to an AI, that sentence goes into the file where it belongs.

Then Use It

Save the files into folders matching the tree, wherever your work already lives. Then one habit does the rest, you open the top folder and ask from there instead of starting a fresh chat from nowhere.

In the Claude desktop app, type @ in the chat box and pick the top folder. In Claude Code, just open the folder and start. In Claude or ChatGPT Projects, make one Project per branch and put the top context file in each.

Once your apps are connected, the same question stops coming back as text to copy around and starts landing as a draft in your Gmail or an event on your calendar. Tools are how your AI stops being useless has that setup, and which apps not to connect.

One Honest Bit

You don't build this in a day, and your first version will have gaps, because nobody can list everything they know about their own work. The gaps show themselves in use. Every time you catch yourself re-explaining something, put that sentence in the file it belongs in and carry on. That is the whole maintenance job, and in a way you are getting your system to learn from you.

And the bounded promise. A structure makes AI genuinely helpful, it does not make it right. You still check the work, same as you would from anyone who started last week.

Where This Goes Next

The best first branch to build properly is the writing one. The folder that writes like you drops straight into the tree above. For what belongs in a context file, the context is what your agent actually knows, and if the word agent still feels like jargon, an AI agent is a chat with a job is the three minute version.