Wabi 2.0 by Eugenia Kuyda: Free Multithreaded AI Assistant

Executive Overview

Eugenia Kuyda, founder of Replika, has officially launched Wabi 2.0, a free multithreaded, multiplayer personal AI assistant that operates as a messaging app and automatically builds custom applications. The launch follows a $20 million pre-seed funding round and positions Wabi as a direct challenger to single-thread assistants like Muse and Instinct. The app is available now on the web for all users, is SOC 2 compliant, and stores most user data in the United States. No token-based pricing or subscription model has been announced, making the tool free at the user interface level.

Deep-Dive Features & Architecture

Wabi 2.0 is built on a multithreaded architecture where each thread is dedicated to a long-term task or goal—for example, a health thread with a weight-tracking app, or a family thread with an app for nearby children's events. When users connect their email, the agent automatically creates a number of initial threads and apps, and users can request additional custom apps on demand. This shift from a social platform for building mini-apps via prompts (in the 2024 version) to an agent that builds apps itself represents a fundamental evolution in the interaction model.

The second key feature is multiplayer design: threads can be shared as group chats, where all participants see shared apps and can query each other's information within a defined scope. In a thread with her partner, Kuyda demonstrates apps for tracking concerts, a daily relationship question, a restaurant list, shared calendar events, and a book list. Kuyda says that single threads in competitors like Muse and Instinct degrade over time due to context accumulation in an infinite conversation, increasing the likelihood of hallucination, context loss, and security control bypass.

• Multithreaded task isolation: each thread maintains its own context, reducing cross-task interference.
• Automatic app generation: the agent builds custom apps based on user goals and email integration.
• Multiplayer group threads: shared threads function as group chats with scoped information access.
• SOC 2 compliance: independent audit of data security controls.
• Transparent personal data requests: all participants are notified when personal information is requested.

Benchmark & Competitive Performance

Wabi does not publish standard benchmark figures, but qualitative comparison with Muse and Instinct reveals architectural differences: the competitors rely on a single conversation thread, while Wabi isolates context into separate threads per task. This design addresses a known problem in AI literature: the longer the conversation, the higher the likelihood of hallucination and loss of tracking. Additionally, the group-sharing feature for threads is not currently offered by competing platforms, according to Kuyda, who stated: "I don't think anyone has solved the multiplayer problem yet."

Industry Impact & Enterprise Adoption

The absence of token-based pricing means that development costs for startups in the Arab world remain zero at the user interface level, but the actual cost shifts to internal large language model consumption, which Wabi has not disclosed. For Arabic text, token efficiency depends on the underlying undisclosed model, and Arabic typically consumes 30-50% more tokens than English in multilingual models. Practical use cases include managing shared family tasks, tracking health habits, and planning social activities—areas that consume significant context. Enterprises evaluating Wabi should note the SOC 2 compliance and US data storage, which may satisfy certain regulatory requirements, but the lack of published benchmarks and pricing transparency warrants caution for large-scale deployments.

Conclusion

Wabi 2.0 represents a significant architectural bet: that multithreaded, multiplayer design can solve the context accumulation and hallucination problems that plague single-thread assistants. With $20 million in pre-seed funding, SOC 2 compliance, and a free-to-use model, Eugenia Kuyda's new venture is positioned to challenge established players. However, the absence of published benchmarks, unclear long-term pricing, and unconfirmed Arabic support mean that users and enterprises should verify capabilities through official channels before committing to production workflows. The app is available now on the web, and its evolution will be closely watched as a test case for multiplayer AI assistance.

Media Source: TechCrunch AI | Fact Verification & Analysis: AI Tools Oasis