Wabi, the startup that let anyone build apps from prompts, is no longer selling the prompt. It is selling the message.
The company announced this week that it is shifting toward an AI messenger experience. The twist: this messenger still builds apps to help you get things done. The founder, Eugenia Kuyda, describes Wabi 2.0 as a “personal agent that does stuff for you and builds the interface you need in the moment.”
Kuyda previously founded Replika, the AI companion startup. Her new project wants to compete with AI agents like Meta’s Muse and Instinct, which combine conversation and productivity tasks in a single interface. The pitch is that Wabi 2.0 delivers both talk and tap.
From Prompt to Messenger
The original version of Wabi let users generate apps from natural language requests. Wabi 2.0 keeps that capability but wraps it in a conversational shell. Instead of typing a command and getting a result, users talk to an agent that responds with text and builds a UI on demand.
Kuyda explains the shift on X. She argues that chat-only agents suffer from a pileup problem. “The more you use a chat-only agent, the worse the experience gets,” she writes. “Your history and ongoing tasks end up buried in one endless scroll.”
Her solution is a hybrid model. “People don’t just want to type-type-type; they want to tap-tap-tap, scroll, and look,” she continues. “They love what agents can do, but they also like software.” She closes with a prediction: “The most powerful agent will be the one that combines both.”
What the Agent Builds
The demo examples center on practical, personal apps. In a thread about health, Kuyda suggests users could build a calorie tracker, a weightlifting log, and a weight tracker. These would appear as accessible interfaces tied to the conversation.
Another thread focuses on family. Here, the agent could create a calendar of kids’ activities, a language learning app, and an app that tracks local kid-friendly events. The idea is that each request triggers a custom interface rather than a wall of text.
“So far, personal agents have been one person, one chat, a stream of one-off tasks,” Kuyda writes. “We think the future is your agent, your apps, and your people in one place, running the parts of life that matter most, together.”
The Invite Code Problem
Wabi 2.0 is not open yet. The app is available only by invite code, which is being handed out on X. That limits who can test the new experience.
The timing matters. OpenAI is also thinking about how apps will work inside generalized chatbot interfaces. Today, OpenAI announced at its DevDay event an update to ChatGPT’s plug-ins. Those plug-ins will let apps offer interactive experiences within ChatGPT, with the app’s name pinned in a more visible spot on ChatGPT’s sidebar.
Both moves sit in the same space: companies figuring out how AI assistants should handle tasks.
Why the Pivot Makes Sense
The pivot is a bet on friction. Kuyda’s argument is that chat-only agents accumulate history and bury ongoing tasks in a single scroll. A tap-based interface lets users see what they asked for and what they got.
She also points to a broader consumer desire. People want software, not just conversation. The agent should do both.
The invite-only rollout means Wabi 2.0 will start small. That could be a smart move. It lets the company gather early feedback before opening the floodgates.
OpenAI’s Plug-In Approach
OpenAI’s plug-in announcement is worth watching. It is a parallel path to embedding apps inside chatbots. Wabi 2.0’s UI-on-demand model is distinct, but both are answers to the same question: how do you make AI assistants more useful?
The plug-ins announced at DevDay treat apps as extensions of the chat interface, with the app’s name pinned in a more visible spot on ChatGPT’s sidebar. Wabi 2.0 takes a different route: the agent builds the interface itself.
What to Watch Next
The next few months will show whether Wabi 2.0 catches on. The invite code limits exposure, which means growth will depend on word-of-mouth from early testers.
OpenAI’s plug-in approach is worth watching too. It is a parallel path to embedding apps inside chatbots. Wabi 2.0’s UI-on-demand model is distinct, but both are answers to the same question: how do you make AI assistants more useful?
Key Facts Box
- Founder: Eugenia Kuyda
- Previous startup: Replika, the AI companion
- New product: Wabi 2.0, an AI messenger with app-building capabilities
- Availability: Invite code only, distributed on X
- OpenAI comparison: ChatGPT plug-ins announced at DevDay
- Example apps: Calorie tracker, weightlifting log, weight tracker, kids’ activity calendar, language learning app
The pivot keeps Wabi’s core skill — building apps from prompts — while answering a real user complaint about chat-only agents. Kuyda’s framing of the endless scroll problem is part of her argument for why Wabi 2.0 exists.
Whether users agree is the open question. The invite code means only a fraction of people will see it soon. That fraction will determine whether the tap-tap-tap model gains traction.
The OpenAI comparison is interesting. Both approaches solve the same problem — making AI assistants more productive — but through different means. Wabi builds the UI; OpenAI pins the app name. Neither has to win for the category to expand.
For now, Wabi 2.0 is a working prototype with a clear thesis. The thesis is that conversation plus interface is better than conversation alone. Whether that holds up at scale is the test ahead.
Source material: “AI-powered app maker Wabi pivots to a messaging experience,” TechCrunch.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.