Buttons Beat Chat
Chat is the dominant interface for AI in 2026. We think that's a mistake. Here's why named, reusable actions beat freeform text boxes for serious work.
The dominant interface for AI in 2026 is a text box. You type, the AI responds, you type again. It's the same interaction model as email, or instant messaging, or SMS. We took the most powerful reasoning technology in human history and gave it the interface of a group chat.
This is a problem worth solving.
The Repetition Tax
Chat interfaces optimize for flexibility — you can say anything, ask anything. But flexibility has a cost: repetition. Because you can say anything, you end up saying the same things, over and over.
"Summarize this." "Now make it shorter." "Make it more formal." "Add bullet points." "Remove the first section."
How many times have you typed variations of these instructions today? How many times this week? Each one is a repeated prompt — information you're regenerating from scratch instead of capturing in a stable, reusable form.
This isn't just annoying. It's a signal that the interface doesn't understand the domain. Mature tools name their important operations. Primitive tools make you describe them every time.
Presets Are Not a New Idea
The music industry figured this out decades ago. A mixing board doesn't make you describe what "more bass" means every time. You turn a knob. The adjustment is encoded, named, and immediately accessible.
DAWs have presets. Photoshop has actions. Video editors have effects. Every mature creative tool eventually develops a vocabulary of named, reusable operations — because professionals stop wanting to describe the same action every time they want to perform it.
AI tooling is five years behind.
What One Click Does
In Kontext, AI presets are named operations on the selected content:
- Expand — broaden the idea, find its edges
- Deepen — go more specific, more rigorous, more detailed
- Switch — apply a different perspective, framework, or lens
- Challenge — surface the weaknesses, the counterarguments, the risks
One click. No typing. The context is already there; the AI knows what to do.
This isn't just a convenience improvement. It changes the relationship with AI output. When you can cycle through Expand, Deepen, Switch, and Challenge in thirty seconds, you stop treating the first response as the answer. You start treating it as a draft — one point in a space of possible responses worth exploring.
That's a fundamentally different way of working with AI. More like a sculptor than a typist.
The Interface Is the Constraint
Chat is a good interface for discovery — for open-ended questions where you don't know what you want yet. But for structured, repeatable cognitive work, it's a constraint in disguise. It makes the same operations more expensive every time you perform them.
Buttons are not a step backward. They're a sign that someone has thought hard enough about a domain to name its important operations. Kontext is a bet that AI-assisted thinking has structure worth naming.
The text box isn't going anywhere. But it shouldn't be the only thing we have.