The meeting your AI books with you every morning

Why It Works

In a chat, you are the one moving the information. You go and get the thing, you paste it in, you ask the question, you carry the answer somewhere else. Run the same model on a schedule instead and it does the fetching, so what reaches you in the morning is the finished thing and nobody had to ask it anything.

What Goes In It

Five things, and the test for each is the same. Would you have gone looking for this yourself today, and would you have been slightly annoyed about it.

1. 1Today's meetings, and what you need before each one

   Have I met them, what did we say last time, and what did I promise them, read out of your own mail.

2. 2The replies you owe, already drafted

   The threads sitting on you and the chases you are waiting on, each with a draft in your drafts folder and nothing sent.

3. 3The notes you wrote away from your desk

   Whatever you put on paper or a tablet yesterday, picked up and put where it belongs.

4. 4The numbers you would not otherwise look at

   Mine is what the running watch recorded overnight. Yours might be the invoices nobody has paid.

5. 5A line that says the whole thing ran

   Every job that fed it, named, with the time it ran, so a dead job never reads as a quiet morning.

Step One, Set It Up

Here from the video? This is the line. It asks which setup you are in first, because the Claude app with your accounts connected and Claude Code run this differently, then it interviews you about what you go looking for each morning, tells you plainly what it can and cannot reach yet, writes the standing instruction, and runs it once against today so you see tomorrow's brief before you commit to anything.

Step Two, Teach It What You Actually Read

After a week you will know three things without trying, the sections you read every day, the ones you skip, and the things you still went and looked up yourself. That is the whole review, so tell it. The brief that survives is the one shaped by what you actually do in it, not by what sounded useful on day one.

Step Three, The Morning Schedule

A brief you have to ask for is another job, so it goes on a schedule and it lands where you were already going to look. The calendar is the first thing most people open, which is why mine arrives as a fifteen minute event at the start of the day, and if your day starts in your inbox, send it there instead. A brief in a place you have to remember to visit is a brief you stop reading in a fortnight.

One Honest Bit

Writing the instruction is the quick part, and it is not the work. The work is everything it assumes, your accounts connected, somewhere for the AI to read about how you operate, and something running it while you are asleep. Mine took weeks.

The failure worth guarding against is not a broken brief, it is a confident one. If it cannot reach your mail it will not say so, it will write a shorter brief that looks perfectly fine. That is why the rules make it name what it could not get to, and why every section gets written even when it is empty, so a quiet day and a broken job never look the same.

Where This Goes Next

If your brief came back generic, the missing piece is context, and the context is what your agent actually knows is the background file that fixes it. If it came back short because it could not reach anything, tools are how your AI stops being useless is which accounts to connect first. And the folder structure is where all of this lives, one place your AI works out of rather than a chat you start again each time.