Wabi started as a tool that let anyone build apps with prompts. Now it’s becoming something closer to a messaging platform, one that can still spin up apps on demand but positions the conversation itself as the primary interface. That’s a meaningful shift, and it reflects a broader tension playing out across the AI industry right now.
As TechCrunch reported, the company this week launched Wabi 2.0, framing it as a “personal agent that does stuff for you and builds the interface you need in the moment.” That’s the language of founder Eugenia Kuyda, who previously built Replika, the AI companion app. She knows something about keeping users engaged inside a conversation over time, and that experience is clearly shaping how Wabi is thinking about this.
The core critique driving the redesign is practical. Kuyda argues that pure chat-based agents get worse as you use them. History piles up, tasks get buried, and navigating back to anything becomes a chore. Her read is that people don’t just want to type at an agent indefinitely. They want to tap, scroll, and interact with structured interfaces. So Wabi 2.0 threads together both: a messaging layer where you talk to the agent, and purpose-built app interfaces that surface inside that thread as needed.
The examples she gave are telling. A health-focused thread might generate a calorie tracker, a weightlifting log, and a weight tracker, all accessible from within that conversation. A family thread might include a kids’ activity calendar, a local events finder, and a language learning app. These aren’t separate downloads or separate sessions. They live inside the conversation that spawned them.
This puts Wabi 2.0 in direct competition with a different category than before. The original Wabi was competing against vibe-coding tools and no-code builders. The new version is going after AI agents that blend chat with productivity, think products like Meta’s AI assistants or whatever OpenAI is building toward with its updated ChatGPT plugin architecture. OpenAI announced at Dev Day this week that ChatGPT plugins will support interactive app experiences embedded inside the chat interface, with app names pinned visibly in the sidebar. The direction of travel is clear across the industry: agents are eating the app layer.
What makes Wabi’s bet interesting is that it’s not just one-person, one-chat. Kuyda’s framing explicitly includes “your people,” suggesting a shared or social layer where the agent, the apps, and the people around you occupy the same space. That’s a more ambitious product vision than most agent builders are pitching right now, and it’s also harder to execute.
The competitive field here includes:
- Meta’s AI agent suite, including Muse and Instinct, which are already reaching scale through existing social surfaces
- OpenAI’s evolving ChatGPT with embedded app experiences
- Specialist productivity agents like Notion AI and Linear’s AI features that handle structured tasks within established workflows
- Other no-code and vibe-coding platforms that are also trying to stay relevant as AI-generated interfaces become table stakes
Wabi 2.0 is currently invite-only, with codes being distributed on X. There’s no public pricing or availability timeline yet. For developers and founders watching the agent space, the product is worth tracking, not because it has won anything yet, but because the design argument it’s making, that the best agent is one that combines conversation with real software, is the right argument. Whether Wabi can build that before better-resourced competitors get there is the actual question.
