The previous half finished two things that weren't particularly comfortable: who gets to take your thinking process, and how the battlefield of data security is moving from the "vault" to the "stream."
Seeing all that clearly is useful. But if it stops at seeing clearly, a person tends to become defensive, tense, guarded about everything.
So the second half turns a different direction: given that process information already exists, what do I do with it.
This is the first article of that half. It is about something I have done for a long time without ever giving it a name.
1. I had been doing the same thing all along, without noticing
Lay the timeline out, and I have done several things that look completely unrelated.
I wrote a plugin that keeps a ledger of my conversations with AI.
I record my meetings and have AI walk through them afterward.
I use these records to look back at myself, and to see other people.
For a long time I thought these were three different habits.
Only later did I realize they are the same action: turning a process that was invisible into something visible.
2. This method is actually very old
That action has a more formal name: the scientific method.
The core loop is four steps — observe → hypothesize → verify → correct.
It is not new at all; it is how humanity has come to understand the world for centuries. What actually changes is never the method, but the instrument of observation.
The naked eye could not see the detail in the night sky, so the telescope appeared, and then astronomy.
The naked eye could not see cells, so the microscope appeared, and then biology.
The naked eye could not see elementary particles, so the collider appeared, and then high-energy physics.
Every upgrade of the instrument of observation has produced a leap in understanding.
And now a new object of observation has appeared: our own thinking process.
What makes this object special? It could never be observed before. Your thoughts happen inside your head and scatter the moment they're done — even you cannot reconstruct them. What you can recall is always the conclusion, never the path.
Until now, when there are tools that record the thinking process.
3. Doing the work: keep the process
Start with the line about work.
Anyone who writes code has had this experience: a bug blocks you for a long time, and in the end the answer turns out to have appeared in some earlier attempt — you just didn't recognize it then.
The problem is that you cannot recall how you were thinking at the time. You remember "I tried a few approaches," but not why you ruled one of them out — and that ruled-out reason was often the answer.
I've kept these conversations, and I have gone back to them.
Reading back produces a strange feeling: you see an earlier version of yourself circling the same problem, hesitating, walking into a dead end. You can see plainly what that version was missing.
That feeling itself is not useful. What is useful is the next step: you start to notice your own patterns.
How do you break a large problem into smaller ones?
How do you choose between approaches?
At which step did you start drifting?
Which of the AI's objections actually changed your direction?
These questions used to be answerable only by feel. Now there is material.
And once you can see your own patterns, you can change them.
4. Being a person: see the process of communication too
Now the line about people.
Most friction at work isn't a problem of logic. It's that two people's receiving frequencies don't line up.
One colleague is impatient — goes straight for the conclusion, can't stand preamble. You, by habit, lay out the whole context before giving the conclusion. So before you reach the point, they are already impatient; you think they're being disrespectful, they think you're too slow.
This kind of thing cannot be solved by improvising — because you don't know where the problem is. You just think "this person is hard to talk to."
I record my meetings and have AI walk through them afterward.
The point is not to keep an archive. It is to see each person's pattern: who tends to interrupt under what conditions, who habitually circles, which topic detonates on contact, and what manner of speaking keeps the other person listening.
Once you see it, responding becomes simple. With a blunt person, lead with the conclusion. With someone who circles, give more patience. With someone sensitive to losing control, sync progress more often.
This is not flattery. It is adaptation.
And adaptation costs far less than trying to change a person.
5. The two lines are the same thing
Work and people look like two directions — one inward, one outward.
But they use one set of movements: record the process, revisit the process, discover your own or someone else's patterns, and then adjust.
It is also exactly what I do when writing code.
Without version control, you don't know how the code became what it is.
Without debug logs, you cannot locate the root of a bug.
Without design documents, no one later understands why the decision was made.
Process information is the foundation of all analysis, optimization, and collaboration.
Technical skill improves on it. So does the ability to work with people. We used to see only results. For the first time, we can see the process itself.
6. Foundation, fulcrum, lever
Let me lay out the three layers.
Process information is the foundation — without it, nothing after it is possible.
Observation is the fulcrum — with the foundation in place, you still need an action to lever it. That action is observation: stop, look, think.
AI is the lever — it drives the cost of observation almost to zero. Reconstructing a meeting used to depend on memory and scattered notes; now you record it, have AI walk through it, and the patterns surface on their own.
The fulcrum itself does not need to be large, but it determines what you can move.
7. The most valuable thing in the AI era
There are two kinds of people who use AI.
One cares only about output: code written faster, documents written faster, information found faster. In their hands, AI is a faster keyboard.
The other asks one more question: how exactly was I thinking just now?
That question looks unproductive. But what it points at is exactly what AI cannot replace — knowing how your own thinking happens, and knowing how other people's thinking happens.
The first makes you get things right. The second makes you get along with people.
And both begin with the same action: write it down, and then observe.
This is the third article of The AI Thought Quartet, Part II: Sovereignty and Practice. There is one more in the Practice half, about the more fundamental thing underneath this method — how you see the world.
Top comments (0)