We Make to Know / Chee Ann

The original version of this post was six sentences. I wrote it on Day 1 of a shipping experiment before I could explain what I'd noticed. Five days later, I can.

On Day 1, I built a company website I'd been "planning" for a year. Sat down with Claude Code, said build me one, and went through four completely different design directions in a single session.

I didn't plan four directions. I didn't brief anyone. I just kept going no, no, closer, no, that.

Four pivots. With a freelancer that's weeks of back-and-forth and the particular guilt of asking someone to redo something you can't articulate. With an agent the iteration was basically free, so I could be honest. I didn't know what I wanted. I found out by looking at things that weren't it.

That evening I wrote a short post. Six sentences. The gist was: the old loop was think → build. The new loop is intent → artifact → understanding. I published it because it felt true, even though I couldn't fully explain it yet.

Then it kept happening.

Day 2, I built a TikTok analytics dashboard. Standard stuff... campaign breakdowns, weekly tracking. Halfway through, while staring at the campaign data, I thought: wait, this is the wrong unit of analysis. The interesting question isn't which campaign is performing. It's which product is about to move.

So I built a scoring engine underneath. Classifies products by velocity, flags what's about to break out versus what's bleeding money.

I didn't plan a scoring engine. It showed up while I was building the dashboard. The artifact told me what the artifact should actually be.

Same evening, I launched cheeann.com. The one you're reading right now. Three years I'd been "planning" this site. Pick a domain. Open Figma. Stare at a blank artboard. Wonder if I have enough to say. Close the laptop. Repeat next quarter.

I had a folder called "Personal Site Planning" with six documents in it. None of them produced a website.

You know what produced a website? Opening the terminal.

Three years of thinking about building. One evening of building without thinking.

Day 3, I opened the TikTok dashboard to adjust the color palette. Eleven commits later, at 3:24am, the whole thing had a name, a visual identity, and a report generation system I hadn't planned.

I went in to change a color. I came out with a brand.

Day 5, two AI agents I'd been building went from mock data to a live API. 160 real campaigns flowing through commands I'd been testing with simulations.

The agents worked fine on mock data. I already knew they worked. What I didn't know... what I couldn't know until real numbers hit... was whether they mattered.

Within minutes of connecting the real API, the question changed. It went from "does this system function" to "am I building for the API or for the person?" That question didn't come from a planning session. It came from watching real ad spend flow through a system and suddenly seeing the operator on the other end of it.

Here's what I think is actually happening.

When building something takes minutes instead of weeks, you stop trying to get the plan right before you start. You just build it and let it talk back to you. The artifact becomes the thought, not the output of the thought.

I didn't know what my website should look like until I saw four wrong versions. I didn't know the dashboard needed a scoring engine until the dashboard existed without one. I didn't know my site needed to exist until it already did. I didn't know the agent mattered until real data told me so.

Every time, the same pattern: I understood what I was making after I made it.

Six days ago I wrote: "We don't just make what we know. We make to know."

I still think that's right. I just didn't know what it meant until I kept building.