How dead is UI? This week, I’ve read a few blog posts announcing the death of the user interface in the Age of the Agent. The central argument: Intent is the new UI – why bother with app interfaces when you can just type up your intent and send it to an agent?
That may be true in some cases, but there’s a number of reasons this seems like a premature death announcement. A few points:
Articulating intent can be really difficult! Some use cases are clear and succinct (e.g “File an expense report with these receipts”). But often, they’re more vague: “I need to create a shareable document on this topic.” UI lays out a possibility space, builds a mental model, and acts as a conduit to tease out intentions. Not to mention, working through a task is often how the full complexity of an intention becomes clear. The ability to articulate an intent upfront can actually be an advanced expert skill.
Intent isn’t always the goal! Exploring, comparing, confirming – these are all complementary or even at odds with a direct intention. Deciding what takeout to order, choosing between different plane tickets, or finding the right way to visualize data all beg for interactive surfaces that go beyond text. The indirect path isn’t friction, it’s the whole point.
Words are limited tools! There’s a reason the transition from CLI to GUI was so powerful – direct manipulation is often easier than typing. The best mapping for a domain – from music to visual design, and even many programming tasks – doesn’t always pass through natural language. Re-mapping everything to written language can actually make things harder. Or as @NotTuxedoSam pointed out on Twitter, you might type 17 words in Claude just to make a 1-character CSS change.
This isn’t a knock against agents. But either/or is the wrong framing – it’s agent plus UI where the most powerful experiences will happen.