Not a tree

Working with AI brings many surprising moments. Work that seemed like it would take ages appears in front of you almost instantly, and ideas that were once vague become something you can actually try out. But looking at the results, some questions stay with you even longer. Is this really what I wanted to make? What should I have explained more? How much should I hand over, and where should I make the call myself?

At our workshop Prompt Hail Mary, we laid out our own methods and concerns and worked alongside AI. It was a time for people with different experiences to sit in front of the same problem, talk about what good design is, and make things with their own hands. We wanted to hold on to the questions we explored there a little longer. There was simply too much we were curious about to cover in a single meeting.

So six of us came together and started Not a Tree. It is a study group exploring AI and design engineering together, with a particular focus on the possibilities of interface design. We are currently on our second presentation.

The name comes from Gyuyoung's presentation, "Design Is Not a Tree." Behind it is Christopher Alexander's "A City Is Not a Tree," which we also discussed at Prompt Hail Mary. How did an essay about cities end up naming our study group?

The tree Alexander describes is not a tree with leaves, but a structure that branches from large limbs into smaller ones. It is like dividing a city into districts, and those districts into smaller units. In such a structure, two groups are either completely separate or one sits entirely inside the other. But the relationships among people living in a city don't divide so neatly. To describe these overlapping relationships, he proposes a structure called the "semilattice."

The essay features a traffic light and a newsstand outside a drugstore. People waiting for the light read the newspapers, and sometimes buy one. The newsstand is part of the drugstore, yet it is also part of the time people spend waiting to cross the street. Relationships that stay hidden when you classify the drugstore, the traffic light and the sidewalk separately come to light when you look at what actually happens there. Alexander warned against cutting away these relationships of everyday life in order to fit a city into a structure that is easy to organize. That doesn't mean overlapping everything is good. The point is to find the overlaps that real life needs.

What if we brought this idea into our own work?

To get work done, we have to break problems apart. We separate planning, design and development, and we look at color, type and layout on their own. Yet while we are actually making something, all of these keep influencing one another. Enlarge the type a little and the whole mood of the screen changes. A constraint discovered during development makes us rethink the original layout. And while weighing ease of implementation and ease of use, we sometimes revise a choice we loved visually.

This is the kind of relationship we want to hold on to in the phrase "Design Is Not a Tree." Drawing on a designer's expertise and intuition, we look at each element one by one, while never losing sight of how they work together. It means looking a little more closely at what is hard to explain with familiar categories alone. Starting from Alexander's thinking on cities, we have widened the question to how we look at design.

Something similar happens when we work with AI. We think we have explained what we want well enough, but the result turns out a little different. And as we explain it again, we sometimes discover that we hadn't clearly decided what we wanted ourselves.

How much judgment is packed into words like "easy to read," "balanced," or "a little calmer"? How far can we explain the sense we have built up through experience? What can be turned into rules, and what has to be seen and judged directly?

So at Not a Tree, we read theories of color and typography, of how people perceive form, and of the interaction between people and computers, and we hold up the ideas of thinkers and experts from across the humanities against the work we do today. We want to ask one another where our familiar principles came from, and whether they still fit the problems we face now.

Then we turn what we have read and discussed into small experiments. We might try writing down design intent a little more precisely, or turning the relationship between color and type into executable rules. These could lead to scripts that check accessibility, tools that reduce repeated mistakes, or skills a coding agent can follow. An idea that sounded convincing on paper may not work well once we actually build it. When that happens, we take that gap back with us to read and talk again.

We also built a small device into how we work together. Each of the six members starts with a topic of their own. One person presents each month, and the others connect a concept or question from that presentation to their own work. Even hearing the same talk, each of us will take away something different. Someone might linger on a single example for a long time, while someone else finds their next experiment in a sentence that clashes with their own thinking.

Each time, we plan to record where our thinking stands now, what we took in, where it clashed with our earlier ideas, and what we will try changing next. We are curious how the ideas we started with will change as they pass through everyone else's presentations. Alongside the outcomes, we want to document that process of change as well.

Our first cohort meets every other Thursday for six months at OTP Company. We share what we have read and written together, and bring back what we made individually or as teams to look at it again. Our writing, tools and records of our experiments will build up together on the Not a Tree website.

We are still a small group, only on our second presentation. We are curious ourselves about what shape these six questions will take. We will put rough ideas on the table, build rough tools, and refine them little by little with each other's help. When we come across something interesting along the way, we will share it here too.

We hope you'll follow along!