Let’s talk

Dwell Wild

Health Community Land

ThreshEarlier work of our team
Learn

The ideas behind it

How do architects come up with designs?

Published Oct 19, 2026 · 7 min read

At Thresh Architects, in Tennessee, I come up with a design by starting without the answer and finding it through many tries. In my 2019 master's thesis I wrote down one way of doing it as seven steps for building a model. I built small models from basswood sticks in about 15 minutes each, turned them in every direction, joined the ones that fit and repeated, and added the doors, floors and stairs last. I think trying again finds something new only when each try looks at the design from a new side.

A worktable covered in quick basswood stick models, each made in about 15 minutes with no answer chosen yet, because a design starts without knowing what it will be.

I start a design without knowing what it will be

When you follow the design process properly, "you shouldn't actually know what the thing is going to be in the end," I said in a 2026 talk, The Architecture of Creation. Many of us want to "start off already knowing where we're going," and then draw that idea out and build it.

My thesis says a design worked out by trying "must be discovered and could not possibly have been designed void of process, from the top down," meaning without the tries, drawn whole before any of the work. Drawing it that way "would require the designer to foresee and simulate all the interactions and deviations and discoveries beforehand," and no designer could foresee all of that.

Not knowing is "kind of scary," and the middle of the work brings "a lot of failures." I try to use the process "in everything that I do," and "I've never been let down whenever I use it." My design process is "just the way that I found to do it."

I compared the work to a marble sculptor's and to the saying that "the sculpture is in there already and you just have to uncover it." For me, uncovering a design means trying one option and then the next until the design shows, so "most of the stuff that I design I didn't exactly invent it."

I start a design without the answer because at the start I don't understand the problem well enough, and, as I said in the talk, the design I find by working is usually better than the ideas I began with.

Small stick models, about 15 minutes each, give many starting points

Hands build a small stick model over a cutting mat on a worktable crowded with dozens of finished basswood models, a glue bottle, a knife and a kitchen timer.

My method starts with small models of basswood sticks, each built toward one small shape in about 15 minutes and then set aside. Step 1 is choosing one kind of simple piece, and I chose "linear basswood sticks." Basswood is a soft, light wood. I took the sticks as they were, without questioning them, since that step begins "Don't make it overly complex, yet."

Step 2 is a short list of rules for how two sticks join. Step 3 is what I called a sprint, a small model built toward one small shape I had in mind. I borrowed the word from software, where, in my thesis's words, a sprint is "a short period of development to complete the development of a whole idea," and mine lasted about 15 minutes.

Once I finished a sprint, I set it aside and built a different one, then set that aside and built another.

While building, I let some sticks run longer than the shape needed. They worked like temporary construction lines, the light guide lines a drawing starts from, standing up in three dimensions.

Each small stick model is built toward one small shape in about 15 minutes and then set aside, so I end with several small models, each a possible start for the design.

Turning the models in every direction shows which ones fit together

Three steps with hands and stick models: one turned over and over, two held together to test the fit, and then joined into a new taller form.

Step 4 joins the sprints, after each one is turned over and around to see which ones fit together. It starts by letting go of the small shape each one was made for. Letting go "disrupts the design process, which is good," because it frees you to "discover what the construct is," the construct being the model.

When two sprints fit well, you add or cut sticks until they join into one larger model, and you keep going "until you have connected as many of the original sprints together as you can."

In my thesis I showed what turning does with an elephant. If you draw its outline, then "turn him ninety degrees to the side" and draw it again, the same elephant "now has a very different border." A stick model gets a new outline with each turn too, and each new outline is another chance to see where two models fit.

Turning each small model in every direction shows a different outline with each turn, and the models that fit are joined into a larger one.

Trying a design again finds something new only when each try comes from a new side

In my thesis I traced new ideas to "disruption of our current perspective," seeing the problem from a new side, more than to sketching the same view "over and over and over again to get it exactly right."

Step 5 is "Rinse and Repeat steps 1-4," which means building new sprints, turning them to see them from new sides and joining them again, a little bigger each time. In the talk I described my design process as picking one option and then the next, and "you keep doing that over and over and over and over until you bump into enough of the context to where it starts to render." Context here means the place and everything around the thing being designed, and once the design "starts to render," its form begins to show.

"If you want to innovate, optimize for perspective and iteration," I said in the talk. Iteration means trying again, and perspective is the side each try looks from, or the person who sees it. "I want to get as many different perspectives on the problem as possible. I want to hear all those voices," because from "a singular perspective, I will only ever come up with a solution that solves from that perspective."

I think trying again finds something new only when each try looks at the design from a new side, because, as I said in the talk, from one view "I will only ever come up with a solution that solves from that perspective."

Doors, floors and stairs come last, once the form is set

My method adds the doors, floors and stairs at the end, after a pause to look the finished form over. That pause is Step 6. "Once you reach a point in the process that you like and are happy with, being able to see what it is you would like it to become, freeze the process." Freezing it only stops the work for a while, and you explore the model again "with fresh eyes and an open heart," so you can see it apart from the ideas you began with.

"Push and pull as necessary for additional requirements such as doors, floors and stairs," Step 7 says. Those come "at the end, after the memory and character has been established," which in my method means after the many tries and turns have given the form its shape.

I wrote that working the rooms in from the start, "in a simple and linear fashion," would leave a form with "no memory, no history, and no context." The rooms would be set down in the model in a step-by-step order, with none of the tries and turns behind the form and nothing in it shaped by the place.

Once you like the form, you pause and look it over fresh, and only then add the doors, floors and stairs, so they're fitted into a form whose character is already set.

A finished design is like an old oak, whose shape was made by every year it grew

A gnarled old oak with a broken limb beside a stick model of a three-story frame, where tweezers place a small stair between its floors.

"This final step is the rendering process," Step 7 goes on, "much like taking a photograph of a tree." Rendering there means the finished form. A photo stops one moment, and in a photo of an old oak the shape you see was made by every year the tree grew.

The tree in my thesis was an oak I might plant. I could "even estimate what it will look like," but I could never say exactly what shape it would take 100 years after planting, with each branch and leaf "acting and reacting to the world around it."

Staring at a 100-year-old oak "brings on a sense of timelessness," the thesis says, and "The oak is timeless because it lives in so many times past." As I see it, the oak looks the way it does because each year's growth was shaped by what was around it that year, on top of every year before. A design found through many tries is built the same way, each try added to what the earlier ones made. So I can't draw it before I start, just as I can't draw the oak's shape at 100 years on the day I'd plant it.

I see a finished design as a photo of an old oak, whose shape in that moment was made by every year it grew, the way the design's shape was made by every try before it.

Sources

  1. Zachary Correa, "The Architecture of Creation," Thresh Architects | Wild Dwelling on YouTube (recorded July 23, 2026)
  2. USDA Forest Service, Fire Effects Information System, "Tilia americana"