Replika Founder Eugenia Kuyda Launches Wabi AI Assistant

Eugenia Kuyda launches Wabi, a multi-threaded personal AI assistant that builds custom apps automatically for users.

Threat Times Systems DeskOctober 19, 20262 min read

Illustration: A sleek smartphone rests on a clean white desk, its screen displaying a minimalist interface with abstract geometric shapes organizing themselves into distinct applic — AI-generated illustration: Threat Times

Key points

Wabi launches as a free, downloadable multi-threaded AI assistant.
The platform automatically generates apps and threads based on user data.
This marks a pivot from Wabi’s initial social-platform design.

Eugenia Kuyda, the former CEO of Replika, officially launched her new startup Wabi on Monday. The release introduces a personal AI assistant that functions as a messenger app, allowing users to maintain multiple agentic threads for various tasks and long-term goals.

According to TechCrunch, Wabi is positioned to capitalize on an idea Kuyda presented more than a decade ago. The platform is free and downloadable to anyone today, offering a distinct approach to personal AI assistance compared to competitors like Muse and Instinct.

Evolution from Early Chatbot Concepts

Twelve years ago, Kuyda presented a conversational chatbot called IO at TechCrunch Disrupt. IO acted as a lifestyle concierge, helping users book restaurants and vacations. It was the first product from her startup Luka, which later became Replika.

Kuyda told TechCrunch that IO still worked despite predating modern large language models. She described the early technology as pretty magical but acknowledged it was nowhere near current capabilities. Luka eventually evolved into Replika, the popular AI chatbot companion company.

Kuyda left her role as CEO of Replika last year to found Wabi. The technology has now caught up with her original vision for Luka. This timing allows Wabi to compete in a growing market of personal AI assistants that manage inboxes and find travel deals.

Shift in Product Strategy

The Monday launch represents a pivot for Wabi, though Kuyda describes it as an evolution. When TechCrunch covered Wabi’s initial launch last year, alongside its $20 million pre-seed round, it was described as a social platform. Users could create mini apps with prompts to share with friends.

Over the past year, the team realized users prefer personal software over one-size-fits-all App Store solutions. However, they may not want to spend time building these tools themselves. Wabi 2.0 addresses this by having the agent decide what apps to build.

When users connect their email, the system automatically spins up threads and apps. Users can add to this by requesting custom apps or starting new threads. Kuyda noted that Wabi does what competitors do but adds these apps to help users accomplish more tasks.

What should you do?

Security professionals should monitor the integration capabilities of new AI assistants. Users connecting personal data sources like email to Wabi should review the permissions granted to the application.

Background: Large language models

Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.

Read the full guide: Large Language Models vs Generative AI: The Distinction and Security Implications

What to do and how to stay safe: Wabi

Review the data permissions granted when connecting email accounts to Wabi.
Monitor the custom apps and threads generated by the AI for unexpected behavior.
Check for updates regarding data privacy policies as the platform evolves.
Verify the security settings for any shared apps or threads with other users.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.