Startups

Eugenia Kuyda's 12-Year Vision for Personal AI Finally Arrives With Wabi 2.0

Replika founder launches Wabi, a multi-threaded personal AI assistant that realizes the lifestyle-concierge concept she pitched at TechCrunch Disrupt in 2014.

Last verified:

A 12-Year Bet Pays Off

On October 19, 2014, Eugenia Kuyda stood at TechCrunch Disrupt and demoed IO, a conversational chatbot that promised to serve as a lifestyle concierge—booking restaurants, vacations, and managing other tasks. The technology was functional but limited; Kuyda reflected that it “still worked” and was “pretty magical,” but fell far short of what AI assistants can do today. According to TechCrunch, that early vision for Luka (which would eventually become Replika) has now materialized as Wabi 2.0, launched publicly on October 19, 2026—exactly 12 years later.

Wabi 2.0: Architecture Over Monologue

The new product pivots from Wabi’s original launch concept. Last year, TechCrunch reported that Wabi was framed as a social platform where users manually constructed mini apps via natural-language prompts to share with friends. Wabi 2.0 inverts this model: the AI agent decides what apps to build.

Upon connecting an email account, Wabi automatically generates a set of threads and starter apps—a weight-loss tracker in a health thread, nearby children’s events in a family thread, inbox management in a communications thread. Users can request additional custom apps or create new threads as needed. According to TechCrunch’s interview with Kuyda, this is “doing what Instinct and Muse do, but it also has all these apps, which helps users do more stuff.”

The multi-threaded design directly addresses a known failure mode in long-running single-thread conversational AI. Researchers cited by TechCrunch confirm that as chatbot conversations extend, hallucination risk rises, context tracking degrades, and safety guardrails become less effective. Kuyda frames the problem plainly: “Single thread experiences, if you think about it, get worse over time because it’s just so much context in one endless conversation.”

The Differentiator: Context Compartmentalization

Wabi 2.0’s core wager is that compartmentalizing tasks into separate threads prevents the context-collapse problem that plagues monolithic assistants. Threads remain shareable, allowing users to delegate or collaborate on specific goals without exposing the full conversational history.

This architectural choice reflects lessons learned during Replika’s evolution. After stepping down as CEO last year, Kuyda founded Wabi to test whether the market had matured enough to support what she originally pitched—a personal software platform that breaks free from the “one size fits all” App Store model. The team discovered users wanted personalization without manual app-building friction, which Wabi 2.0 now automates.

Why This Matters

Wabi 2.0 arrives as a cohort of personal AI assistants (Instinct, Muse) are raising significant capital and signing early users. If the multi-thread architecture proves durably superior—both in reducing hallucination and improving user retention over long timescales—it could become a table-stakes feature for the category. The competitive pressure suggests the industry is converging on a core insight: monolithic chat-based assistants may not be the final form for personal AI, and structural solutions to context degradation matter as much as model capability. Teams evaluating personal AI infrastructure should monitor whether Wabi’s thread-per-task model outperforms single-thread designs in real-world retention and error rates.

Frequently Asked Questions

What is Wabi and how does it differ from competing products like Instinct and Muse?

Wabi is a multi-threaded personal AI assistant that auto-generates custom apps and separates tasks into dedicated conversation threads. Unlike single-thread competitors, Wabi's architecture prevents context degradation and information overload as conversations grow longer.

How does Wabi 2.0 differ from Wabi's original launch concept?

The original Wabi was a social platform where users manually built mini apps via prompts to share with others. Wabi 2.0 shifts to agent-driven app generation—the AI automatically creates threads and apps based on user connections (e.g., email inbox management) and requests.

What was IO and how does it relate to Wabi?

IO was the conversational chatbot lifestyle concierge Eugenia Kuyda pitched at TechCrunch Disrupt on October 19, 2014, as the first product from her startup Luka (which later became Replika). Wabi realizes that 12-year-old vision now that LLM technology has matured.

#personal-ai #agents #replika #wabi #eugenia-kuyda