Ve is
Software since 1984 has been built on click trails. The next decade will be built on intent — short, declarative, increasingly literate. Here’s what we mean by it and why we think the curve has bent.
We’ve spent forty years teaching people to talk to computers in mouse moves. Every product is a sequence of buttons; every workflow a path through a menu tree. A user with a goal in mind translates that goal into a click trail and then walks the trail, hoping the software remembered what they were after by the end.
It works. It also bleeds. The translation cost — goal → trail — is the largest tax in software. Most of the productivity software industry exists to *reduce that cost*: keyboard shortcuts, command palettes, shortcuts of shortcuts.
An intent is a goal stated in language: "draft a reply that pushes the meeting", "find the contract we used last quarter for the same arrangement", "schedule the next kickoff before Friday." The intent doesn’t specify the trail; it specifies the destination.
Three things had to happen for intent to become a viable interaction layer. Models had to read intent literately enough to plan against it. Systems had to plumb their own state to a model so the plan could touch real data. And the cost per intent had to fall to within an order of magnitude of the cost of a click. All three happened in the last eighteen months.
Intent does not look like a chatbot. It looks like a thin, always-available surface — a cursor that knows what you’re looking at, a sidebar that drafts in your tone, a notch that tells you what’s about to happen. The intent text is short, declarative, often passed through with no edit at all.
The interesting design problem is *progressive disclosure of disagreement*. The model says what it’s about to do. You either let it run or correct it. Every correction makes the next pass cheaper.
The first computing wave was the GUI. Direct manipulation, mouse, click trail. It made the computer feel personal because you were *running* it.
The second wave is intent. The computer reads what you’re after and proposes. You correct. It learns. It feels personal in a different sense: not because you’re running it, but because it’s reading you.