I Don't Use AI to Replace My Thinking

Core thesis

I use AI for a lot. It routes mail, tracks finance, watches cron jobs, and keeps me from forgetting the obvious stuff.

But I do not want it doing the things whose value comes from me being present. Sometimes I reach for AI because it helps. Sometimes I reach for it because it feels like I should be using it. Those are not the same thing.

The line is not that AI is bad. The line is that some things stop working when I hand them off.

1. Journaling without reflection

I built evening-journal-reflection as a Hermes cron. It ran every night and gave me a clean summary. Then I stopped reading it. That told me enough.

Now I write by hand in a paper agenda.

What I want AI to do here is narrow: remind me, recover old notes, surface recurring themes. I do not want it to write the reflection itself. If it writes the page, I am no longer thinking in the open. I am just approving a summary of myself.

2. Real photos instead of generated ones

I use real photos, not generated images, when the point is a specific moment.

A real photo says somebody chose that frame. Somebody was there. That small fact matters more to me than polish.

3. Concept graphs by hand

Excalidraw is useful when I already know what I am organizing.

When I am still figuring something out, handwriting wins. The pen keeps up with the thought. The keyboard gets in the way. AI can help later by turning the mess into structure, but not at the moment when the mess is still useful.

That is the boundary: support, not substitution.

Where AI helps

This is the part I think I use best when I use it well:

• find the note I wrote yesterday
• pull the related draft or source file
• surface a pattern I keep repeating
• turn a rough thought into an outline
• keep the vault searchable and usable

That is enough. I want AI to support the work, not replace the part that makes the work mine.

The principle

Journaling without solitude is prompting.

The three examples point to the same thing: the friction is part of the product. Slow writing, a real photo, a hand-drawn map of an idea. Remove the friction and you often remove the reason the thing existed in the first place.

What this article is not

• not a claim that AI is useless
• not a productivity framework
• not an argument that analog is always better

It is a line drawn from inside the tooling. I use AI every day. That is exactly why I can be specific about where I do not want it.

The sharper version

AI has always existed in some form. Maybe now it is just better than before, but if you do not know what to do, do not make the AI do it for you. Do it manually one time. If the workflow does not exist yet, it is usually not worth automating.

I did that with the journal reflection. I wrote it by hand first. Only after that did it make sense to let Hermes help by finding old notes and recurring themes.

The rule is simple: if a workflow matters because it forces me to think, I do it by hand first. If it only feels useful because I am afraid of not using the tool, I probably do not need it.