The team had been framing it as a way to recap your browsing and pick up where you left off. But "picking up where you left off" undersold it. The feature didn't just recap. It anticipated. The real value was that it "suggests what to explore next," because that's what the product actually did. One phrase could reshape what the product meant, but it also meant pushing it through reviews where other teams might see it as a risk. The push mattered, and the only way to make it land was to defend the words themselves. That's the thing no one warns you about: in AI products, the copy isn't describing the product. It's deciding what the product is.
For years, technology was experienced through buttons, images, and audio cues—visual and sensory elements that designers arranged for users to navigate. Now the interface has flipped to its near opposite: a blank text field shaped entirely by the person typing into it. Words have become the single asset almost anyone can use to achieve pretty much anything. And that shift didn't just change how we interact with technology. It changed what technology responds to.
Language models are not efficiency tools bolted onto existing workflows. They are an extension of the oldest information technology humans ever invented: language itself. And because language makes up almost the entirety of the experience—the interface, the prompts created by users, and the feature descriptions—onboarding flows live or die by the clarity of a single sentence more than ever. The people choosing those words are not decorating a finished product. They are doing structural work. A feature description in a spec doesn't just communicate what something does; it shapes how a team thinks about it, how engineers scope it, and how users understand it. A name for a capability isn't a label. It's an architectural decision.
So when we talk about the role of words in AI, we're not talking about polish. We're not talking about the copy that gets added at the end, after the real decisions have been made. We're talking about the material the product is made from. Every naming decision, every feature description, every onboarding sentence is a structural choice that determines what the product becomes and how people experience it. The teams that treat language as an afterthought will build products that feel like afterthoughts. The teams that treat writers as architects will build products that know exactly what they are, because the words got there first.