Articles / Will Smith Spaghetti Test: Can AI Design a Usable App?

Nov 15, 2026 | Articles

#community#Development#digital#digitaltransformation#Technology

Will Smith can eat spaghetti now. Can your app?

In March 2023, someone typed “Will Smith eating spaghetti” into one of the first AI video tools. What came back was horrifying: melting faces, spaghetti everywhere except his mouth, and a stock-photo watermark still showing. It went viral for exactly that reason.

Three years on, the same prompt produces something you have to look at twice. The clip became a running joke, then an informal benchmark. Every time a new video model comes out, someone runs it to see how far things have come. The test is simple: does it look real?

I wanted to run the same test on the apps we design. But for an app, looking real isn’t enough. The question is whether it works for the person holding the phone.

Design took the same trip

AI design tools have grown up alongside video. In 2023, GPT Engineer, the side project that became Lovable, let you describe an app and get code back. In 2024, Figma pulled its first AI design feature because it kept producing copies of Apple’s Weather app. Like the watermark, it gave back what it had learned from. By 2025 we had Figma Make, Google Stitch and a fast-growing Lovable. This year, Figma added its own AI agent.

The growth is hard to ignore. Lovable reached $100 million in annual recurring revenue eight months after launch. By June 2026, over 50 million projects had been built on it, with about a million new ones every week. Many of the people building them aren’t designers, and that’s a good thing. More people can make things than ever before.

The speed is real value

For our clients, that speed matters. Instead of waiting weeks for a polished design, people react to something they can click within days. Instead of betting on one idea, you can try five, keep the one that works and bin the rest before real budget is spent. And one clickable screen in the room beats ten different ideas in ten people’s heads.

But speed only helps if what you’re looking at is good.

So I ran the test

I gave Lovable a fair prompt for a running app’s home screen. What came back was dark, polished and believable. Most teams would happily put it in front of users. It also had five problems:

  1. Tap targets too small. “All runs” was 20 points tall and the tabs about 40 by 30. Apple recommends 44 by 44, roughly a fingertip, and 20 points is below even its 28-point minimum.
  2. It ignored the safe area. The date sat jammed against the Dynamic Island.
  3. Nobody checked the logic. It said Monday, but Wednesday and Thursday already had runs logged.
  4. It hid content. The Start Run button sat on top of my last run.
  5. It wasn’t really an iPhone app. Lovable builds web apps styled to look like iOS, so the text ignored the size people set on their phone.

None of these jump out at first glance. All of them would reach real users. Ship that, and the outcome is complaints, not conversions.

What it actually took

I kept going. When I asked Lovable to review the screen against Apple’s guidelines and WCAG 2.2 AA, it fixed one of the five: the tap targets. When I pushed back on the dark look and asked for the latest iOS rules, it changed the font and the tab bar, but the text still wouldn’t scale. The other three it never noticed. I spotted those and fixed them by hand.

AI did the making, every improvement started with someone asking the next question.

AI accelerates. Judgement delivers.

I’m not saying don’t use it. We use AI tools in our discovery work to get a first clickable version in front of people quickly, so the conversation is about something real. Then our designers take it into a proper design system for the real build.

Creating screens will soon take almost no time. Knowing which one is right will be the whole job. That’s still design. It’s just a different part of it.

 

Let’s Talk AI design >>

Move from AI Chaos to AI Clarity

If your organisation is experiencing fragmented pilots, unclear boundaries, regulatory anxiety or difficulty scaling beyond experimentation, it is time to move from chaos to clarity. Download the whitepaper and take the first structured step toward building AI systems that work

Take the Next Step to Delivering a Better Employee Experience Through AI with Vidatec

employees around the desk collaborating

Learn More About our AI Consulting Services

We cut through the AI hype to help you understand what actually works in your specific situation. Drawing from 300+ real-world projects, we help you implement AI in simple, practical steps that actually stick and deliver results.

Watch our 30-Minute Expert-Led AI Webinar

Hear directly from our CEO, Ross Wilson, and CTO, Nick Welch, as they provide a framework to assess your organisation’s AI readiness and outline practical tools and strategies to implement a true human-first AI implementation.

Read More of Our Latest Insights on AI and Employee Experience

Want more practical insights from our tech team? We regularly share real-world lessons from building AI solutions, developing apps, helping businesses transform digitally, and tackling challenges across different industries.

Brands We Work With

Meet the author

Matt Graham

Matt Graham

UI/UX Lead