For a while, using AI as a designer mostly meant asking it to make something faster. I could generate an image, draft some copy, or get help with code. That was useful, but it was still the old workflow with a faster tool added to it.
What interests me now is different. I’m beginning to redesign the workflow itself.
The pace changed before the job title did
The tools are improving at a speed that makes any rigid process feel temporary. A technique that feels advanced today can become ordinary a few weeks later. I don’t think the answer is to chase every release. I think it is to become more adaptable, to understand what should stay consistent even when the tool changes.
For me, those constants are clarity, structure, character, usefulness, and craft. AI can accelerate parts of the process, but it does not decide what the work should communicate, what belongs, or when something finally feels right.
My role is becoming less about producing every piece manually and more about directing a connected system without losing judgment.
I stopped treating AI like a shortcut
Building this portfolio made that shift visible. I used AI to explore editorial illustrations, adjust visual direction, write and refine interface code, test responsive layouts, optimize images, and repair motion that did not behave properly. The first output was rarely the final answer.
The real work happened in the corrections: make the image less like generic 3D, preserve the original skin tones, use the project’s actual color, hide rounded corners without damaging the composition, keep an animated mascot from overlapping content, and make the same idea work on a narrow screen. Each correction carried a design decision.
That is why I don’t see prompting as a replacement for designing. The quality comes from observing closely, naming what is wrong, setting useful constraints, and checking the result in its real context.
Designing the system around the output
I’m also exploring what happens when those decisions become reusable. Instead of solving the same visual problem from zero, I can define a method: how an illustration should use a brand palette, how much detail it should carry, what must remain authentic, how it should behave responsively, and how the finished page should be inspected.
This is where design engineering has become important to me. The interface, visual identity, motion, accessibility, and performance are not separate finishing stages. They affect one another. A beautiful image that loads too slowly changes the experience. A clever animation that hides text or ignores reduced-motion preferences is not finished. A desktop composition that collapses on mobile is still only a partial idea.
Turning visual judgment into a skill
When I first got into vibe coding, I wanted to build a simpler 3D website design tool: describe the site, and there it is. That was the idea before Visual Designer.
I tried to make it using free trials from different AI agents. It quickly became clear that the idea was much more complicated than it sounded. I subscribed to Codex Plus and tried again, but ran into the same problem. A prompt could get me part of the way, but not all the way to a tool I could trust.
That pushed me to narrow the problem. Instead of trying to generate an entire 3D website, I started building a tool focused on better visual structure. That became the custom capability I call ‘Visual Designer’.
I’m building it to carry a more complete visual workflow: establish the intended feeling, define the visual system, create or direct the assets, assemble the page, and inspect the result rather than stopping at a plausible first render.
The interesting part is not automating taste. It is making the reasoning around taste more explicit and repeatable. Why does this composition feel unbalanced? Which detail is carrying the identity? What should remain untouched? Does the visual still work when it moves, loads, or changes size?
The skill is still in development, with plans to commercialize it soon. I’m also thinking about dividing it into more focused versions for real estate, health, and other kinds of websites. For now, building it is teaching me how much of design lives between the visible output and the decisions that created it.
What I’m exploring next
I want to keep working where design, code, and AI overlap: more expressive web experiences, better ways to direct consistent image systems, faster paths from an early concept to a working prototype, and tools that help designers keep control of the result.
I’m not trying to predict exactly what the designer’s role will look like. I’m trying to build a practice that can move with it. The tools will keep changing. The responsibility to notice, decide, edit, and care for the final experience will not.
