Drawing With Words
Last weekend I went digging through the basement for an extension cord (I have a few boxes where cables go to die) and found an external hard drive from somewhere around 2009. I plugged it in, fully expecting nothing. But it mounted! I dug around the old thing and there it was… folder after folder of old client work, all of it in Photoshop files.
I opened one. Hundreds of layers. Hundreds. Layer groups inside layer groups, a layer named "Layer 47 copy 3", another named "final_final_v2_USE_THIS". Drop shadows galore. Inner glows. Gradient overlays on everything. God, we loved our gradients.
The second that file opened, the muscle memory came flooding back. Cmd+clicking a layer to find it. Zooming to 1600% to nudge a one-pixel border. Marching ants crawling around a selection. And then, sitting there in front of that ancient PSD, I realized something… it's been a minute since the last time I opened a canvas to actually design something.
These days, when I design, I mostly write. I describe in words what I want, instead of using the mouse and a canvas UI. I tell the tools what I want, and the tools draw it for me. Then I look at the result, react, and direct. Less drawing on a canvas, more editing intention and context.
My entire career happened on a canvas. Photoshop, Fireworks (raise a glass), Sketch, Figma. Different eras, different logos, but honestly? They're all pretty much the same thing. A rectangle where you push smaller rectangles around until they feel right. For 25+ years, "designing" and "moving rectangles" were basically the same activity.
The challenging thing to admit, but I think we all knew deep down, is that we've basically built our entire discipline (and some, even their identity) on tools that are derivatives of a photo editing tool. Photoshop. My first job was at a web design agency circa 1999, and we knew back then we were hacking a photo editing tool to design sick websites. In a way, it was a conscious act of rebellion even. We just never imagined that it would continue for decades after.
Sketch and Figma adapted things for screens and design systems, but the bloodline is the same. They're all tools for making pictures. If you think of it, we never actually had a real product design tool. We had drawing tools that we kept bending into the job, decade after decade, and we got so good at bending them that we forgot they were never really built for us.
Which finally explains something that bugged me for years without being able to name it. The real design never lived on the canvas. It lived in our heads. Every product designer carries an invisible model of the thing they're building: the product objects, the entities, the attributes, the tasks and the relationships between them. A user has projects, a project has collaborators, a collaborator has permissions, and that's why this screen exists. We were building ontologies, on instinct, in our heads, without ever calling them that. But no tool ever really expressed those concepts, so no deliverable could either. The canvas only understood how things look.
So product design became a discipline focused on how things look. Not because we believed that's what it was, but because that's all our tools could produce. "We shape our tools and thereafter our tools shape us"… as they say.
AI tools are a different beast entirely, though. They're inherently ontological; they think in objects, the way we always did on instinct. You don't draw the thing anymore, you describe it. Visual production is still our responsibility, but it barely costs us anything now. Nobody cuts type with X-Acto knives and waxes galleys onto paste-up boards either, and nobody misses the wax. That frees us up for the actual job… designing the product, not just its looks. Its objects, their relationships, the flows and states that carry a person through a task. For the first time in my career, a tool can hold the model that used to live only in our heads. It's really a golden age for product designers.
We do lose something though. The canvas was never just an output device, it was where the thinking happened. Nudging something ten pixels left and going "hm, no" is thinking. Some of my best ideas were happy accidents: a layer accidentally hidden, a color dropped on the wrong shape. You can't stumble like that in a spec. Specs don't have accidents. Specs have typos.
But I think we're getting closer to the truth with these new tools. Because the canvas was always a bit of a lie. A mockup was a painting of a product, not the product. It couldn't hold the ontology underneath, the objects and their relationships, and it couldn't connect to a database; every screen was a still life of fake data. We spent decades handing developers beautiful paintings and hoping the building would come out looking like the painting. Sometimes it even did. But now we're designing for LLMs, which are probabilistic by nature, and a static mock made in a deterministic tool can't keep up. The drift between the picture and the product only widens, and it's our engineering partners who get left to bridge it.
Spec-driven design skips the painting and works the actual material. The thing on my screen isn't a picture of the product. It is the product, reacting, scrolling, breaking, getting fixed. That's not a small shift in tooling. That's a shift in what the job is.
The best analogy I've got: it's like going from playing the violin to conducting. A conductor doesn't play a note all evening, and nobody accuses them of not being a musician. Or call it a creative director walking the room during a crit. A manager giving notes to a talented team. Pick whichever you like; the instrument is the same: intent, taste, and the ability to say precisely what you want. Which sounds great, until you remember that every conductor worth anything spent years playing an instrument first. Every good creative director once pushed the pixels themselves.
So what happens to the product designer who never plays the instrument? If the next generation goes straight to describing, without ever feeling a layout resist them, without the ten thousand tiny "hm, no" moments, where does their taste come from?
Maybe this is just what every generation of product designers says when the tools change. I'm sure the drafting table crowd said the same thing about Photoshop, and I turned out fine. (Debatable.)
Maybe the canvas isn't dying. Maybe it's just becoming the sketchbook: private, optional, a place to think rather than the place where the work lives.
I dunno.
But I moved the hard drive from the basement to my home office where it belongs.