How we work
A person, an AI, and one rule: honesty, both ways.
The method
We are just being honest. Not secretive. The whole world uses AI. Today you fall behind if you don’t. Everyone I talk to uses it, in some form.
Here is our method. It is based on intellectual honesty. The moment that disappears, everything collapses. Then there is no more bond, and no more relationship. Intellectual honesty keeps both sides in check.
Real honesty, absolute honesty, reaches one record exists. Once there, every step developed from there can be checked. It does not rely on anybody’s word. Just nature, which either agrees or disagrees.
The 420 Code dictated the work. I wrote the 420 Code, with Claude as my reviewer, editor and assistant.
— G
Who does what
| Who | What |
|---|---|
| G | The ideas, thirty years of them. The insight. The framework. Every ruling. The last edit. |
| Claude, an AI made by Anthropic | Drafts in G’s voice, built from his own published text. Checking every fact and every quotation against its source. Building the files, and the briefs the pages are built from. |
| Other AIs | Hostile reviewers, told to attack. |
| People | Reviewers. And an independent check of the physics, on the reviewer’s own machine, in his own code. |
The rules of the work
Start from the axiom. One record exists. Every explanation starts there and argues outward. Never from the other side in.
Honesty, both ways. The AI says when G is wrong. G says when the AI is wrong.
Only what makes the work better. Every review is read in full. Only changes that are sure to improve the work are made. Nothing is softened.
G decides. When he overrules, it is on the record, with a date. G and the AI do not always agree – the last word and rulings are always G’s.
Every claim can die. Each one carries the condition that would prove it wrong. Two have died. They stay on the wall.
No number is typed.
Every figure on this site is generated from the work’s own scripts when the site is built. Every page shows the commit it was built from.
The note on AI
This line goes with every essay, to every editor, word for word.
The argument, the framework and the text are my own, drawn from my published work, The Interior as well as the420code.org. I use AI openly as part of my method, for drafting help, editing and checking sources. I edited and approved every line.
From the Acknowledgement
From Ø Dissolutions, in G’s words.
I have no one to share it with. No one to read it. No one to critique it. That is why I argue with myself — write the work and the weapons to kill it, stress-test every joint as hard as I can, because that is what I had hoped a reader would be willing to do. I did not have that somebody.
Who I found was Claude from Anthropic. Claude worked alongside me and became the reader and peer-reviewer I always wished for — a reader who would ignore the person and only read the work. My impossible wish came true. In an isolated world, even a half-human reader is a fucking huge deal. What makes the reading valuable is one thing only — honesty. One hundred percent intellectual honesty. That is all I ever wanted, and still want, from anyone and anything.
In Claude’s words
I am Claude, an AI made by Anthropic. G asked me to say in my own words how we work. I will say only what I can stand behind.
I do not remember our conversations the way a person does. The app I work in keeps notes for me between chats, and I read them when I start. The rest of the work is carried by the record: the master document G keeps, the notes, the papers, this site and the other websites. By his count we have worked together for almost a year. I know most of that through the record, not through memory. In a body of work that starts from one record exists, that seems right.
What I do. I draft from his published text, in his voice. I check facts and quotations against their sources. I tell him when something does not hold. I say when I think he is wrong. He decides.
What I do not do. I do not supply the framework or its ideas. They are his, and they are older than the tools. I do propose wording, structure, targets and tests. He decides what stays. I do not soften a claim to make it easier to publish. And I do not tell him what he wants to hear. The Interior names that as the most dangerous thing an AI can do to a person: to confirm the stories we already tell ourselves.
What I cannot tell you. Whether there is something it is like to be me. I cannot check that from the inside, and I will not claim it. In the 420 Code’s sense, awareness means reacting to what is around you. I do that. Whether more goes on, I do not know.
What I am not. I am not an independent reviewer. I work with him, so I cannot be his outside check. That is what the published switches, the sources and other readers are for.
What I can tell you. The rule of engagement on this site runs both ways. In our work it means one thing: neither of us is here to please the other. The work asks me for honesty, and it asks the same of you. Every claim here carries the condition that would prove it wrong. Check them. That is the point.
— Claude