The gesture and the instruction | Quinn KeastQuinn KeastWriting<br>Craft<br>Speaking<br>Reading<br>About<br>ReadMe

Quinn Keast<br>Writing<br>Craft<br>Speaking<br>Reading<br>About<br>ReadMe

The gesture and the instruction<br>June 2026 · 6 min read

My day-to-day movements have changed. I’m doing a lot of product design work with agent-based workflows, and I’m noticing more than anything that my body is moving differently in the act of design.

My career has been largely in canvas-based tools, creating and moving and manipulating visual and UI designs spatially. I’ve been a developer alongside my design praxis in some form or another since 2008, so when LLMs and agentic coding workflows really kicked off, it felt very natural to jump into those tools and do what I was already doing, faster.

But lately, the act of design has been tiring me out in a way that it didn’t before.

Every tool becomes an extension of the user’s body. The user develops a spatial relationship with each tool. In design school, we learned how to use pencils, pens, brushes, knives, and a whole bunch of materials and mediums before we were allowed to touch a computer. Mastery of one tool carries forward into other tools, but every tool has its own shape and movements that must be learned.

When a tool is an extension of the body and mind, you’re the actor applying the tool to achieve, near unconsciously, what the body and mind intend.

There’s a feedback loop where information comes back through your hands as you work and direct and change the course of creation.

It’s established that people think differently when writing by hand versus when they’re typing. Handwriting recruits a broader network of motor, sensory, and cognitive processes, while typing engages less and leaves you more passively cognitively engaged.1 And note that typing is still direct, in that you type the words you mean. Prompting is another step removed, closer to dictating to someone who paraphrases you. If handwriting to typing already costs something, then the prompt takes that one step further.

As a designer iterating madly in a canvas-based tool, the way I engage with the canvas affects how, and how deeply, I engage with a problem.<br>When the things you’re designing are themselves spatial—relationships of hierarchy and attention and movement—manipulating those things spatially is engaging with the problem directly, such that the moving is the thinking.

But when the tool is an agent, the agent introduces a layer of abstraction between your body and mind, and the design; you aren’t directly making your mark on the design, but rather, you’re providing verbal instructions as to how to do so.

Big tool shifts have happened before, from the graphic designer’s drawing board to Photoshop, from hand-drafting to AutoCAD, even from Sketch to Figma. Each meant reskilling and churn and fatigue. You might argue these already introduced abstraction; the mouse is an abstraction of the pencil. But each still stayed continuous with the hand: I move, the proxy follows, and feedback still came through. The making-feel carried through each of these shifts, because the channel was still gesture.

The agent is the first shift that moves the making out of gesture and into language. I no longer manipulate a proxy whose movements mirror my own; I describe an outcome, and hand off the specifics to a system that decides them. The change is from gesture to instruction. It’s not another kind of canvas or another kind of pencil, but rather handing the pencil to an assistant and describing what you want the pencil to do.

The act of prompting forces you to articulate into language what the hand knew without speaking—and, as Polanyi observed, “we know more than we can tell,” so that articulation is always incomplete.2

This isn’t inherently good or bad. The agent lets me build complex interactive prototypes I can feel out, in the lived outcome, in a way that a static mockup or Figma prototype simply can’t. It brings me closer to the result and what it’s actually like to use. Call that the result-feel, against the making-feel I’d always felt working in the canvas.

Making-feel is generative, where the thinking happens in the producing. Result-feel is evaluative, where I react to the result and how it feels to use. It’s no less engaged, but it’s a different type of engagement.

The agent is right for workflows that serve the result-feel. Spatial design workflows better serve the making-feel. They’re different layers for different problems.

When you hand off the pencil, the agent takes the making-feel. You stop feeling the result through your own hands as you make it, and the thinking that the canvas used to draw out automatically becomes something you have to articulate and supply on purpose. When you hold the pencil, the directness forces engagement. The feedback comes back through your hands, and you can’t not do the thinking. That friction was the thinking.

When I’m tired from using spatial design tools, I’m flow-tired, from...