The Challenges of Design Atrophy

Julio Martínez
10 min readJan 16, 2024

Revisiting a distant problem with new tools

Cobweb image explorations

Every now and then, an old problem resurfaces as a renewed curiosity. I always remember a deceptively simple assignment from design school: we were asked to choose a word and then visually interpret the meaning of the word. I am broadly paraphrasing the brief — it probably had a few more constraints — but that was the gist of it.

I arrived quickly at a word choice. For whatever reason, I had become fascinated with the word atrophy. I come from a family of trained medical practitioners, so I suspect I became aware of the word in its medical context at some untraceable point in the past. English being my second language, at the time I was still coming to terms with these kinds of words and their many imposing consonants, their somewhat somber meanings lurking in the background. I also probably liked the shape of the word, the triangular letterforms at the ends, that tempting “O” in the middle, that ancient “ph” sitting there, still quite foreign to me. Regardless of how I got to it, I picked the word and began sketching ways of “seeing” it.

The memories of the entire process are dim, mostly wrapped in tangled cobwebs, but some aspects remain lively in my recollection. In one of my ideas, I pictured an image of a couch or armchair sagging hyperbolically, as if overtaxed by someone sitting on it for decades and decades. I drew this with graphite and charcoal, aiming to use the softness of the material to allude to the absence of the mass. The results were decent starts, but they were missing the impact I wanted. They felt like just “things” — old couches drawn by hazy hands. They were unconvincing possibilities; plausible scenes that lacked a storytelling resonance.

Studies for melting armchair

I figured I’d try a photo approach next, to see if I could get a stronger reaction. I wanted to play with fragments of old, decaying couches, then distort and amplify the wrinkles of the fabric. To do this, I needed a good, old couch. The one in my family’s living room was definitely old, but it didn’t have the wrinkly, malleable surface I imagined manipulating. I needed other options, so I headed to the Salvation Army in San Francisco’s Mission district, hoping to find a more compelling subject. I walked up and down their modest furniture aisle, taking quick reference images with my handed-down 35mm film camera. I photographed all their seat cushions and some stray linens. On the way home, meandering about, I remember seeing an upholstered armchair on the curb, set out as a Freebie for any takers. I snapped a photo of that, too, but it was more torn than wilted, more suggestive of trauma than atrophy.

The photo manipulation of the couches did not work as well, for reasons I cannot precisely recall—though I assume they had to do with my limitations at the time in both skill and RAM. So, I moved on to try other experiments. At the time, my family owned a shoe store in the Mission’s Excelsior district. We had a large basement, with piles of fixtures and furnishings left over from the previous owner, a family that had operated the store for decades before we bought the place in 1996.

Within that vast, concrete cave, I set out to find additional source imagery. I photographed some of our old chairs, some wooden tables, some benches with tears in the upholstery; I shot our out-of-usage shelving, edges of slat walls from the seventies, fragments of leftover all-purpose boards, and even a 1930s-era cash register we still had down there for some reason. I ventured into some of the less-traveled crevices, the nooks and crannies of the space, out back towards our rear office and then towards the front, underneath the sidewalk cellar doors where none of us rarely went. I found cobwebs, rusted surfaces, and lots and lots of decay. I snapped three rolls of film and giddily headed out to a nearby Walgreens to have the images developed and printed.

A sampling of photo resources taken during the exploration

While this process seems endless as I think back on it, only a couple of the resulting composites survive in my hard drives now. I have no visual remnants, no tangible clues as to what I did with most of those photos. Still, I must admit that the remaining few scraps hold a certain charm for me; they show an early hint toward effectiveness that is easier for me to appreciate now than I probably did then. There was perhaps more there than I knew what to do with. Sometimes, it takes a while — decades even — to understand how to work with what you have.

Ultimately, I returned to drawing lines by hand and with the Adobe Illustrator pen tool. This return to basics has been a predominant aspect of my process for a long time: when all else fails, I tend to go back to drawing, making simple marks, and going from there. At some point during this part of the project, I arrived at a clearer concept: I wanted to make a brain made of cobwebs. This felt right as a way to visualize atrophy. I sketched a few variations, editing clusters of these frayed lines over and over. I tried angled lines; I tried duplicating them mechanically, I tried looser approaches. Some of these files still reside in the deeper reaches of my digital archives, and they suggest a fair amount of time was devoted to this line of exploration.

Of course, I don’t really remember exactly where I ended up, which iteration I turned in, or what grade I may have gotten. But I do remember truly enjoying making those drawings, getting lost in the intricacy of that webwork. I also remember simply feeling better about the outcome. It still didn’t quite match my “vision” yet, but it was intriguing. Besides, the end-of-semester reviews were approaching soon. Fully satisfied or not, I needed to move on.

Study for Brain / Cobweb idea, drawn in Adobe Illustrator, circa version 7.0

Many months later, in my first job out of school, I had a project manager once tell me, as he saw me trying to meet a deadline with frantically mixed results, “…this is beyond your experience, but not beyond your capabilities.” I appreciated that sentiment, both at that very moment and to this very day. It applied to that afternoon’s task, whatever it may have been, but it also reminded me of the bigger picture, of the widening horizon. It reminded me of my atrophy image explorations and the many other projects in school where my intent hadn’t exactly aligned with what my hands could do.

Though this feeling recedes as a career progresses, it never entirely disappears. It was particularly frustrating in school because it was a new kind of angst, a new kind of failure. It’s hard as a student to understand that this is simply what happens when you shift from unconscious incompetence into conscious incompetence. But it’s easier to live with it in hindsight. It’s an unavoidable companion in this line of work, a sign that growth is happening. Ira Glass talks about this distance between vision and execution that haunts us all, shadowing every one of our early pursuits.

I returned to this assignment recently — some ghosts never truly subside. Over the last few months, OpenAI made a splash by releasing DALL-E, and the world was flooded with keyword prompts and vibrant imagery made of wanderlust. Discussions arose all over about the future of this specific subdivision of our species (i.e., image-makers), and pleas to remember the humans trapped beneath all these machinations. Lawsuits ensued, panic and turmoil quickly followed. I stayed away from the tools at first, indignant at their overreach, but resistance was futile. After all, during these last few months, A.I. image generators have swiftly gone from being a ”dreamy projection to an ambient menace and perpetual sales pitch.”

I dug into the tools to see what they could do with that old, atrophied concept. I played with simple prompts at first, starting with just typing in “atrophy,” which was useless. That much open-endedness seemed to stump the machines. I then tried being more and more specific: “a brain decaying,” “a brain made out of cobwebs in a basement,” “a human brain drawing made out of frail cobwebs in a wooden frame,” etc. The generators kept rendering the brain too fully, so I tried to hone in on that aspect and went on and on about the need to “make the brain less obvious,” “draw a brain made out of thin cobweb lines,” or “add more decay.” I had some intriguing results a few variations later, but most refused to stray from the fleshy brain. They remained stubbornly literal, unattentive to the metaphor.

A sampling of A.I. generated images, showing selected results from OpenAI’s DALL-E,’s DreamStudio, and Midjourney

Are they interesting studies, though? As with most early forays into these new experiments, the verdict is: Hmm. Well, maybe. They certainly seem more finished than anything I could have approached twenty years ago—or today, frankly. They were undoubtedly faster to make and easier to elaborate on. This optimized vetting process is quite powerful — discarding an idea wouldn’t take weeks and digressive bus rides around town. You can Google some prompts and say ‘yay or nay’ to an idea before lunchtime.

And yet, of course, I still appreciate the coarseness of the original studies. The A.I. images are all MCU, but here, I was hoping to make an indie flick. Despite my prodding, I could not get the brain to be quite so overt — in my head, it was always more of a trace, a ghost of a brain that had already mostly dematerialized. I wanted an image that depicted something that used to be there, something that was no longer present, something with its liveliness pressed out, diminished through time. The fully volumetric brains of A.I., all fleshy, almost living, seemed to belie this intent.

With more time and more attention to prompting, I am sure a more satisfying result could be made, but eventually, I stepped away from the depths of the cave-shaped rabbit hole. This was a curious tangent, not a full-on real project at this time. No, the machines would not easily automate these kinds of challenges. Not now, anyway. I do look forward to finding ways soon in which these tools can benefit designers’ day-to-day workflows. In fact, I am already enjoying the slow removal of the grunt work and the ease of finding an FPO image or a placeholder vector. I am thrilled that I may never have to draw a clipping path with the pen tool, point by point, ever again, not to mention the myriad productivity enhancements soon to become more and more of a standard.

But those benefits come at different stages of the process. Within the ideation phase, the tools are a bit of a mirage. They end up taking away a crucial part of the process: the wrong turns, the failed attempts. They obscure the basic premise behind our design profession’s learn-by-doing ethos. An interaction with LLMs only feeds their own algorithmic hunger, the insights landing mostly on their offshore server farms. The tools leave the human designer—the human learner—at a loss. They leave out the possibility of discovery, of fluctuation, of error-turned-insight. They can only regurgitate what already exists; they can’t extract non-existent, vague notions from your head.

The outcome of this particular project, twenty years after design school, is immaterial. The final solution has been lost to the vagaries of time and the constant shuffle of computer updates. What I do remember, though, is working through the problem. I remember thinking about it, being challenged by it, cursing my decision to choose that word. I remember the chats with my professor about how the exploration was going, the bus ride to the second-hand store, the photos around the basement of my family’s shoe store, the surprisingly generous feedback from a classmate about my couch drawings. These frayed bits of memory and conjecture have stayed with me, and I’ve built on them over time.

So, indeed, these image-generation tools are quite strong, but it’s good to be wary of the complacency they might spawn. Their shortcuts in productivity are impressive, but they take away the memory of the process, the mental track record of making an image. They take away the learning opportunities, removing the possibility that our own neural synapses might be strengthened as we make attempt after attempt. Their prompt-and-wait process generates remarkable results, but they also produce anonymous, foreign outcomes—distant images that leave a stagnant, unsettling kind of atrophy in their wake.



Julio Martínez

Creative Director, Educator, and Illustrator in San Francisco, California. Born in México City.