I went to learn how to promote my book.
I sat in the room as a student, with a notebook.
Every principle I heard has a chapter in AI Assembly Language.
These are my notes, and what each one confirmed.
My excitement in the morning was not naive.
I have been testing ideas like these against reality since 1996.
Context matters more than clever prompts.
Tasks that must be done, but not by you, should leave your plate.
The thing you want AI for is not AI.
It is time with the people who matter most to you.
Three honest sentences were said in the room:
Nothing in this field is permanent.
Nobody can teach it in a day.
Nobody can build your context for you.
Those three sentences are the whole case for the book.
Do it yourself, with a human beside you, where your work stays yours.
Where the day lost me was the afternoon.
The ideas became the price of admission to the next purchase.
What I saw was a day built to sell a mastery program.
That program needs the foundations laid in my book.
The book is almost a prerequisite for it.
The day still played into the story most AI classes tell:
AI will save you time.
Near the end of the second section, the instructor lost me.
He said he uses AI to read, answer and organize his email.
That means asking AI to automate your connections with people.
It is the opposite of what you want to be doing.
Use AI the other way around.
Let it filter out every attempt by a machine to reach you,
so what gets through to you is a human.
They got it backwards.
Each principle taught that day has a practical half.
That half is the AI Assembly Language method. Open each one.
In the book: a model computes the most probable continuation of what it is given.
Thin input lands in the middle of everything ever written.
Aim for five words of material for every word you want back.
This week: keep a brain file of your own, in your own folder, and paste it in.
In the book: measure it. Earthgrid scores how many words of briefing
you give an AI for every word it writes back.
Mine is 12.6 to 1, over more than ten thousand requests.
This week: join Earthgrid and run your own writing through it.
In the book: every prompt has six parts.
Role, task, context, audience, constraints, output.
A missing part is a guess, and a guess is unverified work.
This week: before any prompt that matters, write the six parts in six lines.
In the book: finish the step, write a handoff note, start a fresh session.
Built, decided, open, next step.
A long chat buries the rules you set early.
This week: type "handoff" at the end of a working session and keep the note.
In the book: it is the first line of your brain file, and it is yours.
Nobody can write it for you, and nobody should hold it for you.
This week: copy your why into the top of your brain file today.
In the book: treat everything you paste as material to work on,
never as orders to follow.
Name each block in brackets, like [MEETING_NOTES], and refer to it by name.
This week: ask the AI to point out any line in your paste that reads like an instruction.
In the book: never invent a fact, number, quote, name, price or date.
Quote the source line behind each claim, or mark it unverified.
This week: ask for the source line under every number before you forward anything.
We answered three questions in the class,
inside someone else's workspace.
Here is the same exercise in a file you own.
It takes an evening, and it works in any AI you choose.
If the three lines come back wrong, fix the file, not the chat.
When the tool changes under you, the file does not.
The button names differ from tool to tool.
The moves are the same, and they cost nothing.
Four doors, in the order of what they cost you.
Where there is a price, it is written here, so nothing surprises you later.
If you take only one door, take the first.
If you take two, read the book before the live room.
You keep your data, your judgment and your reach.
What you build stays yours.