Maintaining Taste in the Age of Slop

The other day I was doing some research in the ‘screen time app’ niche to see what sort of options (both established as well as newer) were out there. Pretty typical for the most part, but then I started noticing some…peculiar ones:

Now of course, it shouldn’t be surprising seeing stuff like this, creating cheap knockoffs of successful products is a tale as old as time. But it got me thinking: in the age of AI it’s kind of crazy that people are still doing this. I mean, we live in a world where in software, the time between having an idea and it being real is basically instant. Where the mechanical work of building features is no longer the bottleneck, and where all that really remains is the human aspect: design, UX, taste.

Five years ago, the primary thing that mattered to me when choosing a software product was its featureset. And usually it was the bigger, more established company that had the more mature featureset, and thus, provided more value to me as a customer. But now when I look at the software I use day-to-day, I find that more and more I’m prioritizing how it feels over what it does. To give a few examples:

• Notion: Doesn’t have any features you can’t find in Google Docs. No one was asking for another note-taking app. But every corner of the UI is so polished that it just feels better to use.
• Zen Browser: Basically just a Firefox reskin, but its so clean/customizable/UX-focused that I got hooked as soon as I tried it.

And within the screen time niche, the canonical example of clean design is Opal. And the problem with trying to rip them off is that while you can steal the UI, you can’t really steal the ‘taste’. Because taste is more than just a component’s CSS or a particular mechanic; it’s a reflection of the entire brand and the people behind it, and it only works if it’s clear that those people truly care about what they’re putting out.

While building Island, thinking about how to balance tastefulness with AI-accelerated development has always been top of mind. And I think for us, it boils down to one main thing: we treat our agents like SWEs, and not Designers or PMs. All of our product development and UI/UX is entirely human-led, and we only prompt AI to execute on decisions we’ve already made.

One example that comes to mind is the app’s background water texture. It’s 200+ lines of shader language code and 100% AI-written, and I never bothered to understand the math behind it. Because it makes zero difference to the end-user experience.

HOWEVER, it certainly wasn’t just 1 prompt and done, but rather involved multiple revisions, tweaking colors, intentionally making the pixels isometric instead of square, as well as hand-tuning 35 other noise parameters. Because these are the kind of improvements that make an incremental, yet noticeable difference in experience.

The beauty of AI-assisted development is that you don’t have to be an experienced graphics programmer to create things like this. With just a bit of design sense, what would’ve taken a pro days can now be conjured at the speed of thought.

I’ve been going on about taste for a while now, but of course it’s not a 1-step recipe for success. Even in B2C, differentiation is essential to stand out and carving a niche for yourself, and it’s the aspect in which we saw the biggest opportunity for innovation.

Most screen time trackers are a variation on the same theme: a wall that shows up when you open an app and either makes you wait or locks you out. And once you get frustrated enough or get walled out of a legitimate use case (eg. responding to a message), you uninstall.

We wanted to take a different approach with Island. You already have the desire to reduce your screen time, and rather than being forceful or nannying you, Island serves to amplify that desire by tying your habits to something tangible: a little digital farm on an island. And the more you stay under your screen time limit, the more your farm prospers.