The Ground Continues to Shift
The ground under design work keeps moving, and it'll be a while before it settles.
There is no doubt that designers are in the middle of a transition. When I went to design school in the early 1990s, we were still learning traditional production techniques: drawing on vellum with hairline Rapidographs, cutting rubyliths, rubbing down halftones, and gluing photostats. The industry itself was in the midst of a transition from physical techniques like these to personal computer-based ones brought on by the Mac, Pagemaker, and PostScript.
My production teacher, the very eagle-eyed Bonnie Russell, would look at our mechanicals with a loupe—yes, a loupe!—and mark any slip of the hand. I’d often receive assignments back dotted with red circles.
However, school was the last time I pasted up anything by hand. After I graduated, the graphic design world had already settled on doing things digitally. I applied the same detail-orientation to my Illustrator and QuarkXpress files instead, so Ms. Russell’s scrutiny wasn’t wasted.
The point being there was a decade-long transition. Designers learned these newer digital production processes as they were being proven to work, as they were becoming the standard.
Nearly four years after the ChatGPT moment (GPT-3.5 in November 2022), and over a year after Lovable took off, designers are wrestling with some existential questions that are beyond techniques. When we step back and look at what designers actually do beyond pixel-pushing, AI is eating at those tasks. Some of it is tedious and we don’t want to do them anymore, like tagging concepts in interview transcripts. But sometimes, the tedium is how we internalize and get better. I would not be as detail-oriented or as much of a pixel fucker if it weren’t for Ms. Russell back in design school.
This transition, this unsettled moment causes a form of grief. We mourn what we’re losing, but we’re also feeling stuck. Jack Maguire believes it’s because there’s no fixed endpoint to grieve toward. When a person dies, he points out, the loss is finite and you eventually adjust to it. AI displacement offers no such resting point: retrain into whatever looks safe now and you may watch it automated inside two years. Workers, he writes, are being asked to “accept a process rather than an outcome.”
You can see the same instability in what we even call the work. Sarah Gibbons at Nielsen Norman Group points out that “AI design” has already forked into four different jobs—using AI in your workflow, designing AI products, designing for agents, shaping how a model behaves—and that everyone says the phrase while picturing something else entirely. Even those categories come with an expiration date. She figures the window to build rare expertise in any of them closes within a year, once the rest of the field catches up.
And the titles keep multiplying. Nicole Alexandra Michaelis catalogs the new business cards—Agentic UX Architect, Trust Designer, AI Design Consultant—and I’ll admit I’m wary of the whole genre. I’ve watched titles and labels proliferate into pretentiousness before. The constant she points to is: the strategic judgment underneath is the same craft it has always been, whatever fancy title we print on the card.
Even “developer” has come loose. When a veteran Mac journalist like Jason Snell can ship a working app in an afternoon with Claude Code—”ugly and incomplete,” by his own description, but working—the word stops meaning what it used to. He found the code was the easy part; envisioning and deciding the app was the actual work. The skill that defined the role turned out not to be the role at all.
So if even the definitions and the titles keep moving, what do you actually stand on?
The closest thing to an answer came from Amber Bouabdallah’s piece. Her Salesforce team set out to design how designers master AI and found there was no curriculum to build: no fixed competence to train people toward, because the tools arrive faster than any best practice can be written down. What worked instead was smaller and simpler—designers watching each other work, catching the workarounds someone tries when the tool behaves unpredictably. She says that peer layer “isn’t the bridge. It’s the ground.” There’s no permanent training to cross over to, so the people around you become the thing you depend on.
That answers the question I keep hearing from designers on my team, the one underneath the title anxiety: what’s the safe thing to become? Maybe there isn’t one. The durable thing was never a role or a tool or a name on a card. It’s the practice you carry between them: your judgment and the specific way you bend a tool toward how you already think.
I’ll leave you with a quote from an interview with Tommy Geoco I linked to a few weeks ago (edited for clarity):
It is chaotic. And I think there’s a lot of value in the people, the teams, the individuals who are just sharing their workflow. And I think it’s important to remember that’s a good activity right now. And you don’t have to attach the language to it like, “Hey, here’s my workflow. You should do this.” Strike should from the vocabulary right now. And just compare. Compare workflows, compare shapes. Because we will eventually converge around some better practices around this stuff, but we don’t know what that is yet. And so right now, the best thing you can do is if you want to get involved in this type of work to just try to find workflows that are good for you and spend some percentage of time on the familiar stuff that gets the work done. And then some percentage of the time, make room for yourself to explore a workflow. We are going to converge around stuff. If you have the luxury and you don’t have to get involved yet, you can wait. And there probably will be a concentration of, “These are the better workflows.”
What I’m Consuming
Sweet Jeebus, MacOS 27 Golden Gate Removes the Dumb Icons From Menu Items. macOS 26 Tahoe stuffed an icon next to every menu item, and John Gruber was far from alone in finding the result cluttered and inscrutable. Golden Gate quietly pulls them back out and rewrites the Human Interface Guidelines to push for sparing, purposeful use. A small reversal, but a reassuring signal about the design direction in Cupertino. (John Gruber / Daring Fireball)
From the Collection: Emigre Font Development Files. The Letterform Archive has digitized Emigre’s complete font development files, a world’s first for a digital foundry’s process work, and Eve Scarborough reflects on what the marked proofs, revisions, and reader mail reveal. The files catch type design mid-transition between analog and digital, back when Rudy VanderLans and Zuzana Licko were chasing the promise of the early Macintosh. VanderLans’s aside about how design debate has scattered across too many online channels to follow is a quiet bonus. (Eve Scarborough / Letterform Archive)
“Music venues are the engine of the creative industries”: lessons from the graphic ephemera of our lost clubs and pubs. The V&A’s Lost Music Venues exhibition treats flyers, tickets, setlists, and disco balls as cultural records worth saving, documenting the independent clubs and pubs that have vanished, especially since Covid. Curator Harriet Reed argues these rooms are community infrastructure, not just nostalgia. The graphic ephemera turns out to carry the memory of a whole creative ecosystem. (Paul Moore / It’s Nice That)
AI and Libraries, Archives, and Museums, Loosely Coupled. Dan Cohen makes the case that cultural heritage institutions don’t have to choose between ignoring AI and surrendering their collections to mass training. His alternative is Anthropic’s Model Context Protocol: a non-extractive bridge that lets libraries, archives, and museums connect their collections to AI tools on their own terms, with ground-truth checks and local-model deployments. It’s the rare AI-and-institutions piece that offers a concrete, controllable path instead of a warning. (Dan Cohen)
Startups are science. Jonathan Yagel, a self-described English major who found measurement distasteful, recounts how Eric Ries’s The Lean Startup rewired his thinking: a real startup’s first job is learning under extreme uncertainty, long before growth. His fix was to run marketing like the scientific method, actually writing down the hypothesis and reviewing the results afterward, which almost nobody does. A conversational read on treating early-stage work as structured inquiry rather than guesswork. (Jonathan Yagel / Age of Intelligence)