Wabi 2.0 Becomes an AI Messenger That Builds Interfaces on Demand
Wabi is becoming an AI messenger that can still make apps to help users get things done. TechCrunch The company calls the release Wabi 2.0.
Previously, Wabi let anyone build apps from written prompts. Its founder is Eugenia Kuyda, who previously founded the AI companion startup Replika.
Kuyda describes Wabi 2.0 as "a personal agent that does stuff for you and builds the interface you need in the moment." In practice, the system creates a small interface on demand when a user asks the agent to do a task.
That framing places Wabi 2.0 against AI agents that combine chat and productivity tasks, rather than against vibe-coding tools, which generate software from prompts.
The new Wabi 2.0 app is currently only available by invite code distributed on X.
Wabi's official website describes the product as "A new kind of messenger that gets things done." Wabi The site carries the tagline "More of what you want to do. Less of what you have to." It says "Let Wabi handle your next group trip." Example uses listed are Group trip, Bills & subs, School, Group pet, Games, Book club, Dinner plan and Live music.
The broader context here is the shift from persistent applications to transient interfaces, or temporary screens built for one job. Instead of installing a dedicated app for each coordination problem and learning its data model, which is how it organizes information, the user states intent in conversation and the agent assembles the interface for that task alone. Messaging fits that pattern. It already holds identity, presence and group membership.
In my view, the technical bet is less about generation than about state, or shared information that must stay current. Anyone can render a form. The hard part is shared, changeable data across a group chat that stays consistent while members join late, change votes, split costs or drop out. Add tool use and permissions, or what the agent can access and do, and the challenge compounds. That is where agent messengers succeed or stall.
Looking at what this means for builders, the pivot clarifies the interface question. Vibe-coding tools optimize for artifact creation, or producing lasting software. Agent messengers optimize for task completion. The generated UI is disposable, like a receipt. Conversation is durable. For tech-literate users, that tradeoff will feel familiar from chatops and command palettes, where a short typed command pulls up a tool inside chat. The interface appears when needed, then recedes.
Worth flagging is the human factor around coordination. Coordinating a group trip or splitting bills remains socially awkward software. I have watched my own children ignore full-featured planning apps and settle logistics in a thread with screenshots and IOUs. If Wabi can keep coordination inside the thread where the social pressure already lives, it removes adoption friction. That is a modest goal. Modest goals often ship.
