June 16, 2026
AI Can Bring Back the Old, Weird Interfaces We Loved
From Winamp skins to vibe-designed interfaces, AI is reopening the door to software that feels custom and fun.

By Matthew Stephens
7 min read
Now that I'm up and running with BADS, I find myself exploring so many different beautiful and unique design directions. And I'm building prototypes that have 5–10 different on-brand themes that my client can choose from. And I find myself wondering why do I need to eliminate them? What if the user could choose? Or create their own? It's just a natural language prompt away, if the design system is good enough.
It reminds me of the old days designing Winamp skins, building weird little Flash websites, and customizing Myspace profiles. My digital life was aspirational, anything felt possible. Recreating this scene from Hackers was my ultimate goal in life:
I just feel like, for a while, software got pretty quiet.
Not quiet in the sense of being simple. Just from a UI standpoint. Visually and emotionally, it all converged so quickly.
The same clean white canvases. Rounded rectangles. Soft shadows. Tasteful gradients. Maybe a Corporate Memphis illustration, if you're lucky.
Everything became usable. A lot of it also became interchangeable.
That was not an accident. The last decade of product design was shaped by systems, scale, accessibility, and platform consistency. Design systems helped companies ship faster. Component libraries reduced chaos. Brand guidelines made products feel coherent. Accessibility standards pushed interfaces toward greater clarity. These were real gains.
But something was lost along the way.
Software used to feel more personal.
I will say, I do have a personal bias here. I built a website, a community, and a company that catered to helping people customize their digital experiences.
But that personalization mattered to us back then.
Changing the interface made the computer feel less like a machine I had been assigned and more like a space I could shape. I could make my music player feel futuristic, super weird, or deeply specific to whatever TV show I was into that week. It gave me a sense of authorship before I had the language for design.
A lot of early digital design was like that. It was personal before it was professional. It was expressive before it was optimized. It was sometimes ugly, almost always impractical, but most importantly, it was fun.
That enjoyment is easy to dismiss as nostalgia, but I think it points to something real: people want the digital tools they use every day to reflect something about them.
I believe, the future of software is going to be shaped by that desire. And I believe there's an opportunity for designers to make a career out of it.
AI is not just going to make software smarter. It is going to make software more personal, more fun.
What Does the Future of Personalization Look Like?
Personalization has always been difficult because meaningful variation creates complexity.
The more a product can change, the more things there are to design, test, support, and maintain. That is why most customization options have been shallow. They offer the feeling of personalization without letting the product change too much.
AI changes the equation.
A modern interface does not have to rely only on static themes or manually designed skins. It can generate variation within constraints. It can interpret intent. It can respond to context. It can preserve the functional structure of a product while changing the emotional and visual layer around it.
That is the key shift.
The future of personalization is not total chaos. It is not every user randomly generating unreadable interfaces. It is controlled adaptation.
A product could allow significant personalization while still following rules:
- Text must remain readable.
- Hierarchy must stay clear.
- Critical actions must remain visible.
- Accessibility settings must be respected.
- Navigation must stay predictable.
- Brand behavior must remain intact.
The system must adapt to device, context, and user needs.
This is where AI becomes useful for interface design. Not as a replacement for design judgment, but as a way to make flexible systems practical.
In the old skinning era, personalization was fragile because the skin did not understand the software underneath it.
In the next era, the skin can be aware of the system. It can know what is content, what is navigation, what is primary, what is dangerous, what is decorative, and what must never be obscured.
That unlocks a much more powerful version of customization.
Vibe Designing Is The New Customization Interface
Most people do not want to manually adjust typography scales, layout density, contrast ratios, color tokens, motion settings, icon styles, and component states.
Designers might enjoy that level of control. Most users will not.
But people do know how they want software to feel.
They can say:
- "Make this feel calmer."
- "Make this dashboard less corporate."
- "Make this workspace feel like a cozy notebook."
- "Make this planning tool more playful."
- "Make this interface easier for me to focus in."
- "Make this feel like a trading desk."
- "Make this writing app feel like an old paperback."
That is where vibe designing becomes important.
Vibe designing gives people a natural language interface for personalization. Instead of making users become interface designers, it lets them describe the mood, intention, or working environment they want. The system can then translate that request into design decisions.
That translation is the design challenge.
A good vibe-designed interface is not just a mood board applied to software. It has to understand what should change and what should stay stable. It has to know the difference between atmosphere and function. It has to preserve the parts of the experience that create trust.
The best version of this is subtle but powerful.
- A budgeting app could feel less punitive and more supportive.
- A project management app could feel less like a corporate dashboard and more like a studio wall.
- A health app could become calmer when someone is reviewing sensitive information.
- A creative tool could become more expansive during brainstorming and more focused during production.
- A learning app could adapt its tone and density based on confidence, age, or skill level.
These changes are not just aesthetic. They affect how people feel while using the product.
And how people feel while using a product affects whether they keep using it.
Design Systems Become Guardrails, Not Cages
If personalized software is the future, design systems will have to evolve.
They will still matter. In fact, they will matter more.
But they will need to shift from defining only fixed outputs to defining ranges of acceptable variation.
Today, many design systems answer questions like:
- What does a primary button look like?
- What is the type scale?
- What colors are approved?
- How does a modal behave?
- What spacing values are allowed?
Those questions will still matter. But generative interfaces introduce new ones.
- How much can the product's personality flex?
- Which elements are allowed to change?
- Which elements must remain stable?
- What kinds of visual metaphors fit the brand?
- What accessibility constraints are nonnegotiable?
- How does the system evaluate whether a generated interface is still usable?
- What forms of user expression create delight, and what forms create risk?
The design system becomes less like a box of finished parts and more like a grammar for generating coherent experiences.
That changes the role of designers.
Designers will still design screens, flows, and components. But increasingly, they will also design the rules that allow many versions of an interface to exist. They will define constraints, examples, prompt patterns, evaluation criteria, and quality thresholds.
The work shifts from creating one perfect interface to creating the conditions for many good interfaces.
That requires taste.
It requires judgment.
It requires understanding where flexibility improves the experience and where it damages trust.
AI can generate options. It cannot decide what a product should stand for. It cannot fully understand the social, cultural, business, and ethical implications of every design choice. That remains design work.
The craft does not disappear.
It moves up a level.
Personalized Software Is Also An Accessibility Opportunity
Personalization is often treated as an aesthetic preference. But it can also be an accessibility strategy.
Different people need different interfaces.
Some need larger text. Some need reduced motion. Some need higher contrast. Some need simpler layouts. Some need more structure. Some need less visual noise. Some need language that is clearer, warmer, or more direct. Some need workflows broken into smaller steps.
Today, accessibility settings are often separated from mainstream personalization. They live in settings panels. They are treated as accommodations rather than part of the core experience.
AI could help close that gap.
If software can adapt to personal taste, it should also adapt to personal ability, comfort, and cognitive load. The same systems that let one person make a workspace feel cozy could help another person reduce overwhelm. The same generative layer that changes visual tone could make information easier to parse.
That does not remove the need for accessibility standards. It makes them more important.
Personalized software should start from accessible defaults, then allow responsible adaptation. The user should be able to shape the experience without accidentally making it unusable for themselves or others.
This is one of the strongest arguments for designers staying deeply involved in AI-driven personalization.
Without design judgment, personalization becomes novelty.
With design judgment, it can become a path toward more inclusive software.
How Taste Will Evolve
There is a lazy version of this future where everyone generates endless themes and most of them are bad.
That will happen.
But it does not mean design taste becomes less important. It means taste becomes more important at the system level.
- Someone has to decide what good variation looks like.
- Someone has to know which parts of an interface can flex and which parts should never move.
- Someone has to protect readability, accessibility, hierarchy, and trust.
- Someone has to know when delight becomes distraction.
- Someone has to make sure personalization serves the user, not just the novelty of the technology.
The best designers will not simply make screens. They will shape taste engines.
They will define the prompts, constraints, examples, tokens, and evaluation criteria that allow software to generate many expressions without losing coherence.
That is a different kind of design craft, but it is still craft.
Maybe it is closer to gardening than architecture. You are not placing every leaf. You are shaping the conditions for growth, then pruning aggressively.
Basically, Software Can Get Weird Again
AI has raised the floor drastically for design, but there's an opportunity to raise the ceiling as well. And I think that's what we're missing in the current discourse. This shift can make software weird again.
Not unusable weird. Not hostile weird. Specific weird. Human weird.