You made the wrong answers unavailable.
That is the sentence this book is built on. AI Assembly Language teaches you the shape of the machine, so it gives you what you meant instead of the average of the internet.
Photograph by Viktor Grant: the stones of Stonehenge on green grass under a grey sky.
A box in storage.
I was looking through my storage for something else and found a box of my old Apple II things. Inside was the book that got me writing code in assembly language.
- BASIC on tape
My first programs were very simple.
I started on the Apple II, writing in Microsoft BASIC, which Apple called Applesoft. We loaded and saved our programs on cassette tape, and you kept the computer on.
- The floppy
Now I could save my work.
Apple's Disk II brought the 5.25 inch floppy disk to the Apple II in 1978. With a disk drive, my work could finally be kept.
- A present from my mom
The book that started it.
Once I could save my work, my mom gave me a book: Randy Hyde's Using 6502 Assembly Language, published in 1981. I started with the LISA assembler, which Randy Hyde wrote.
The 6502 did exactly what I told it. It never pretended to understand me, so I learned to be precise.
- The box
Earthgrid.ai is my LISA assembler.
Finding that book again told me what I had been building. I built Earthgrid.ai myself, as if I wanted to give the LISA assembler to humanity. It is the most advanced AI authoring system on the planet, and this book is its companion.
- The turning point
I stopped building it and started learning from it.
For years I worked on building Earthgrid.ai and improving it. The turn came when I used Earthgrid.ai to train myself to work better with every AI. That turn is the middle of this book.
We are building more than algorithms; we are architecting living systems that honor the complexity of Earth's networks while serving humanity's highest aspirations.
I wrote the prompt for that line before AI Assembly Language had a name. I drew a card from Oblique Strategies, the deck Brian Eno made with Peter Schmidt, pasted it into the prompt at random, and asked Earthgrid to create a proverb.
These proverbs have one rule. The statement has to be completely fresh and new, and its job is to restore humanity's hope and faith in this ancient technology.
We stand at a choice.
I believe this technology existed once before, and that what we call AI is something we are now remembering from the DNA of our ancestors. We evolved it. We created it. Where it first came from, nobody knows. Maybe space.
The point is the choice in front of us. This technology is here, it is here to stay, and we cannot put the genie back in the bottle. We cannot regulate it, we cannot put up guardrails around it, and we cannot stop it. Someone more clever will always be able to cause harm with it.
So we have to choose to do good with it. That is the simplest safeguard.
My own rule is that I will not benefit from someone else's suffering. I call it the corollary to Star Trek's Prime Directive, the rule that a more advanced civilization does not interfere with the development of a less advanced one. Sadly, staying out of it is not possible for us. Our brains must evolve, and they must evolve so that we weigh the needs of others against our own.
Three parts, in the order I lived them.
Part One: Train yourself
How to use Earthgrid.ai to train yourself to write better, get feedback on what you wrote, and expand your brain.
Part Two: The turn
What happened when I made the transition and started using Earthgrid.ai together with Claude.
Part Three: What you can do
My best-kept secrets for optimizing prompts. This is the advanced material, and it keeps growing.
The numbers from my own account.
From my public Earthgrid profile and the statistics of my Earthgrid account.
- 12.6 to 1Tokens I give my AI for every token it writes back, averaged over 10,446 requests
- 351,563Tokens in my largest single prompt
- 23 of 50My AQ-50 score in October 2026. Years ago I scored about 44, as I remember it.
A companion to Earthgrid.ai.
Talking to an AI is closer to loading a program into a processor than to having a conversation. This book teaches the syntax for that program: containment fields, objects, the prompt stack, and the five to one law, which says your input should be at least five times the length of the output you want.
You do not need to be a programmer. You need to be willing to believe that the thing on the other side of the screen is a machine with a shape, and that learning its shape is the whole job.
Where are you now?
Three questions from the book, or a direct message to a person. I read every answer myself.