Wabi 2.0 turns its app builder into an AI messenger

Wabi relaunched on September 29, 2026, turning its prompt-based app builder into an AI agent people reach through chat. TechCrunch reports that the new version, Wabi 2.0, still builds small apps on demand, but now inside an ongoing conversation instead of a separate app-building screen.

Founder Eugenia Kuyda told TechCrunch the goal is a "personal agent that does stuff for you and builds the interface you need in the moment." She argues chat-only agents get worse the longer people use them: "Your history and ongoing tasks end up buried in one endless scroll."

What changed

Wabi now sorts a user's tasks into separate threads by topic, such as health, family, or work, instead of one long chat. Inside any thread, the agent can still generate a small working app on the spot, such as a calorie tracker or a shared family calendar, SuperpowerDaily reports.

RuntimeWire reports that the underlying mini-app technology has existed in Wabi's iOS app since a June 20 update, version 2.0.4. This week's launch repositions that same technology around chat and shared tasks. It is not an entirely new product.

Access is invite-only, with codes shared over X. RuntimeWire says the app also offers a paid tier, Wabi Max, at $20 a month.

Money and competition

RuntimeWire reports Wabi raised $20 million in pre-seed funding in November 2025. The round was led by venture firm a16z, which described Wabi as "a new kind of messenger that makes apps and gets things done."

TechCrunch places Wabi in the same market as Meta's Muse and Instinct, another AI agent startup now valued at $10 billion. Kuyda's bet is that combining chat and generated software beats either approach alone. "The most powerful agent will be the one that combines both," she told TechCrunch. RuntimeWire frames the relaunch the same way: less a new product than a change in distribution strategy, moving Wabi away from solo app creation and toward shared, social use.

What this means for developers

Wabi's pivot is a signal, not just a feature update. Prompt-to-app generation alone was not enough to hold users, so Wabi wrapped it in the retention mechanics of a messaging app instead. If you are building an AI app-generation tool, plan for what happens after the app gets built. A generated app people never reopen is not a business.

The thread-based organization is also worth studying. Sorting ongoing agent tasks by topic, instead of one flat chat log, directly answers a real complaint: long-running agent conversations become unusable as history piles up. Any product that keeps state across many sessions will hit the same wall Kuyda described.

Watch the $10 billion Instinct valuation as a sign of how much capital is chasing this exact idea. A crowded, well-funded field means differentiation will matter more than being first. Wabi's answer, mixing chat with on-demand interfaces, is one bet among several serious ones.

The invite-only rollout is also a deliberate choice, not just a launch-day limit. Distributing codes over X, rather than opening signups to everyone, lets Wabi control how fast its thread-based design gets tested at scale before wider release. If you are planning a similar agent launch, a staged invite list gives you room to fix the rough edges a full public launch would expose immediately.