【Genesis of the AI-Native Company | Day 1】 “Creation of the Earth”

＝＝＝＝＝＝＝＝＝＝

“The earth was without form, and void; and darkness was on the face of the deep. And the Spirit of God was hovering over the face of the waters.” — Genesis 1:2

＝＝＝＝＝＝＝＝＝＝

“In the beginning, there was a ‘skeleton.’ The skeleton was with the AI. The skeleton was the very ‘place’ where the AI moved.” — Genesis of the AI-Native Company, Chapter 1, Section 2

📖 Genesis of the AI-Native Company, Chapter 1, Section 2:

Day 0 was when I birthed a paper. I placed the concept of an “AI-Native Company” into this world as a term.

Before building, create the ground

The day after I decided to create an AI company, I realized something. The paper was done. A precise concept already existed. However, there was no “foundation.”

Just as you need to prepare the ground before building a house, an “AI-Native Company” needs a “place” to operate. No matter how excellent an AI tool is, it will never function without a “place” to move. I realized I had to design that “place”—in other words, the skeleton of the organization—first.

The “skeleton” created on Day 1

What I did on this day was not building something using prompts. I prepared three skeletons (structures) for the AI to operate as an “organization.”

1) AI-Native Company Model and New Organizational Structure

Twitter founder Jack Dorsey proposed “From Hierarchy to Intelligence.” I felt a deep resonance with this vision and design philosophy. I deconstructed the design principles of the “self-driving organization” that Dorsey implemented at his company, Square, and reintegrated them into an “AI-Native Company.” I minimized human intervention points and redesigned the loop of decision-making, execution, and verification as an organizational structure.

2) Tool Integration Design Guide

In order for multiple AI agents to cooperate and produce results, I defined the “AI-Native Startup” as a system that prevents AI runaway, loss, or omissions through structure by explicitly documenting role division (who decides what), information transfer (in what order it flows), and human intervention points (where to approve/send back) as a blueprint.

3) Context Vault (AI Context Engine)

1) and 2) correspond to “harness engineering,” while 3) corresponds to “context engineering.” As a result of trial and error, I built an “information map” called Kenji Natsumoto's Context Vault on Obsidian.

My thoughts, philosophy, decision-making, habits, experiences, vocabulary, and values regarding me, Kenji Natsumoto—
I have structured all of that "contextual information" and designed it so that an AI can reference and update it at any time.

Because these information sources are in place, the AI will not misread my thinking or direction, and by understanding my intent, it will be able to design the construction of an AI-Native Company more accurately.

Before the prompt, there is structure.

When people say they "use" AI, most think about prompts. They ask, "How should I phrase this so the AI accurately understands my intent?"
However, to operate AI as an organization, building "context engineering" and "harness engineering" comes first. No matter how excellent the prompts you write are, an AI organization will not function without a skeleton.

Those who pursue efficiency simply look forgood prompts.
Those who pursue effectiveness focus first oncreating a "skeleton".

Preparing the earth. That is, creating the "skeleton."

That is what I worked on for Day 1. On the next day, Day 2, "light" will be kindled upon this foundation—the day will come to choose the decision-maker for the AI-Native Company.

📖 "Genesis of an AI-Native Company" is a true story series about the founding of an AI-native company, which I am telling by drawing parallels to the Old Testament's "Book of Genesis."