Originally published 2022 · Updated 2026 — Steven Tyrrell / Tyrrell Creative

In 2022, I wrote that artificial intelligence had us standing on the cusp of a tectonic shift, and that once the tools matured, A.I. would transform creative work like nothing before it. I hedged about timelines and wondered aloud what it would mean for design and art.

I’m revisiting that piece now because enough time has passed to do something most predictions never get: check the scoreboard. So this is the same argument, but with the benefit of four years of hindsight — and the discomfort of seeing how fast “speculative” became “Tuesday.”

What I said then

The frontier in 2022 was OpenAI’s DALL-E 2, which could turn a written phrase into a startlingly accurate image. Google had just teased a model called Imagen — not yet available, with its eventual public use an open question. I looked at the sample images and called the photorealism a game-changer, then asked whether it might land a death blow on parts of the design industry: stock image libraries, conceptual artists, and a good many photographers.

I used a small thought experiment. If a creative director could say “give me an animated octopus reading a newspaper” and get five high-quality, versatile renditions in seconds, what happens to the illustrator who’d have spent hours on a single version? I worried about a flood of instant digital “art,” the way animation cels went extinct when digital artboards arrived. And I reached for a metaphor I keep coming back to: the blacksmith, master of a trade the world quietly stopped needing, hunting for new work in a changed economy.

Then I asked the uncomfortable questions. Is design next? Is art next? Will an untrained user with no grasp of design principles type a few prompts and get a professionally laid-out page with no mouse clicks at all? My answer was a reluctant “yes.”

What actually happened

The reluctant “yes” turned out to be the right call — faster and more completely than I expected.

Imagen isn’t a teaser anymore. Google’s Imagen 4 family is generally available, and its top tier produces some of the most convincing photorealism on the market — the kind that’s genuinely hard to tell from a photograph. DALL-E didn’t just get better; OpenAI has replaced that line twice, and image generation is now woven natively into the chat tools millions of people already use. Add Midjourney, Adobe Firefly, Ideogram, Google Gemini’s Nano Banana, the open-source Stable Diffusion ecosystem, and a roster of models that didn’t exist when I wrote the original, and “give me five octopuses reading newspapers” is no longer a thought experiment. It’s a free afternoon.

The “untrained user lays out a whole page” prediction landed too. Between Canva’s AI features, design assistants baked into Figma and the Adobe suite, and template engines that generate on-brand layouts from a sentence, a non-designer can now ship something that looks professional (more on that in another post) without touching a single design principle. The mouse clicks I mentioned are increasingly optional.

And the part I most hoped I was wrong about: the market moved exactly as feared at the bottom of the pyramid. The little jobs that kept freelancers fed — book covers, t-shirts, postcards, simple posters, stock-style imagery — are the first to be absorbed. The World Economic Forum’s most recent Future of Jobs report now lists graphic design among the fastest-declining roles of the next five years, a stark reversal from the prior report, which had it growing. The blacksmith metaphor aged uncomfortably well: as far back as 2022, an A.I. image won a fine-art competition at a state fair, and the man who entered it was promptly compared to exactly that — a tradesman whose craft the future was rendering obsolete.

So if you read the 2022 piece as a warning, the warning was accurate.

What I got wrong — and it matters

Where I was too gloomy: I framed this mostly as extinction, and that half of the picture is incomplete.

The same WEF report that flagged design’s decline named UI/UX design one of the fastest-growing roles. The work didn’t vanish so much as it migrated — up the value chain, toward strategy, systems, product thinking, and the parts of the job a prompt can’t fake. Naming, brand architecture, knowing what a client actually means when they can’t articulate it, judging which of five striking options is right rather than merely good — A.I. is still poor at all of it. The designers I see thriving haven’t out-typed the machine. They’ve moved to where the machine can’t follow, and they use it for the grunt work — resizing, variations, background removal, first drafts — so they can spend their hours on judgment.

The other thing I underestimated was how unsettled the legal ground would stay. The U.S. Copyright Office has held that purely A.I.-generated work — with no meaningful human authorship — can’t be copyrighted, which is a real problem for anyone hoping to build a brand on prompts alone. Meanwhile the courts are still sorting out the training data itself: Getty’s landmark case against Stability AI largely failed in a UK ruling, dozens more suits are grinding through U.S. courts, and “where did this model learn to draw” remains a live, expensive question. For a creative shop, that’s not abstract. It’s the difference between an asset you own and an asset you merely possess.

The same advice, sharpened

In 2022 I closed with three things to do. They’ve held up. Here they are, updated for the world we actually got:

1. Keep a professional journal of ideas — and now, of prompts and process. Capture not just concepts but the judgment behind them. The thinking is the part that doesn’t commoditize.
2. Stay current, but go past dabbling. Knowing these tools exist is table stakes now. Build a real working fluency — which models for which jobs, where the legal landmines are, how to fold A.I. into a workflow without surrendering the parts that make the work yours.
3. Be original. Be authentic. This was the closer in 2022 and it’s only truer now that the world is drowning in competent, generic, A.I.-shaped output. The scarce thing isn’t production anymore — it’s a point of view. Nothing replaces authenticity, not even a very good machine.

The blacksmith didn’t survive by making cheaper horseshoes. He survived by becoming something the new economy still needed. With talks of massive layoffs and UBI, how far will the global economy shift and families suffer before it all settles out? The blacksmith ended up on the assembly line at the Ford plant. Generations later, iRobot will be the assembly line. After all, a humanoid named “Flash” from the Chinese company, Honor, broke the world record for the half-marathon a few weeks ago. The tools will keep getting better, cheaper, faster, smarter, and more autonomous.

We can’t forget that we still have an advantage. Humans are incredibly versatile. We originate, create and innovate, and pivot in a way that AI hasn’t demonstrated yet. Non-measurable things that AI cannot learn are still solely ours. Instinct, the gut feeling, the lived experience, losses, suffering, pain and wounding, happiness and joy. They produce perseverance and foster determination to grow and innovate. The call to purpose and legacy and something beyond ourselves will not be supplanted any time soon. It’s time to reinvent ourselves and keep stepping forward into the future.

By the way, what’s a newspaper?

© 2022, 2026 Steven Tyrrell / Tyrrell Creativeell Creative

Original images below were AI prompt generations in 2022.