Civic Innovations

In her excellent book Frostbite, author Nicola Twilley writes about how modern fruits and vegetables have been engineered not for taste, but for shipping. Those grocery store tomatoes? They weren’t bred for flavor—they were bred to survive weeks of cold storage and transport. Their thick skins and firm flesh optimized for logistics, not taste. Twilley raises the unsettling idea that we might already live in a world where nobody remembers what a proper tomato is supposed to taste like.

That idea has been top of mind every time I read something online lately. The internet is rapidly filling up with content generated by large language models (LLMs). It’s optimized for SEO, readability, and safe, generalized tone. It’s (sometimes) technically correct. It’s grammatically clean. And a lot of it is just… bland. Lifeless, unchallenging, eerily competent. Like a tomato at the grocery store that looks perfect until you bite into it.

A lot of what LLMs generate is just empty calories. Content that says nothing, means nothing, is often inaccurate, and exists just to fill the page. The way people use these tools doesn’t tend to produce anything that’s really “creative.” It doesn’t invite deep insight. The goal is speed, volume, and coverage—not originality. The job of an LLM is essentially to remix the statistical average of what’s already been said somewhere else.

And that’s the problem. We are entering an era where a huge and growing percentage of what we read is engineered to be palatable at scale. Content that is made to travel well, not to taste particularly good. It’s the intellectual equivalent of a cold storage tomato. This isn’t necessarily a new problem, but the ubiquity of LLMs has supercharged it.

Some people may not be alarmed by this. Who cares who wrote something if it’s useful? Aren’t humans still in the loop somewhere, editing or prompting? Perhaps. But that’s like arguing there’s still a farmer involved when you buy a flavorless strawberry at the grocery store in January. Technically true, but deeply beside the point.

The real risk is what happens over time. When you grow up eating refrigerated tomatoes, you stop asking for better ones. You don’t miss the flavor of a vine grown tomato because you never knew it existed in the first place. The same goes for writing. If everything starts sounding like the same vaguely helpful listicle or generic blog post, we might forget how distinctive, human voices make us feel.

I don’t think we’re there yet. It’s still possible to find quirky, insightful, deeply human content online. But it’s getting harder and harder to find. And as more content is generated by machines trained to be average, the more interesting stuff will get pushed further and further to the edges. Eventually, we might not even realize it’s gone. Or that it was there at all.

Maybe we don’t need to panic, but we do need to pay attention. If the internet starts to taste like a tomato that’s been optimized to survive months in a shipping container, we should at least be able to name what’s been lost.

Because if we forget what the good stuff tastes like, we’ll stop knowing how to look for it.