Replika's Kuyda opens Wabi to all users: a multi-threaded, shareable AI assistant

Wabi, the new startup from Replika founder Eugenia Kuyda, opened to all users on Monday. The product is a personal AI assistant built as a messaging app, where users keep multiple parallel threads, each dedicated to a task or a long-term goal.

The timing is deliberate. Twelve years to the day earlier, Kuyda stood on the TechCrunch Disrupt stage with IO, a conversational lifestyle concierge built by her then-startup Luka, which later became Replika. That early product worked, she said, but nothing like what large language models can do now.

Wabi 2.0 changes who does the building. Once a user connects an email account, the agent spins up a set of threads and apps on its own, then keeps adding custom ones on request — a weight-loss tracker inside a health thread, a list of nearby children's events inside a family thread.

Where assistants such as Muse and Instinct run everything through a single conversation, Wabi bets on multiple threads and multiple people. Kuyda argues single-thread experiences degrade over time, because the context piles up until users are buried in what the assistant sends them each day.

Threads can be shared and used like group chats. Within a shared chat, users can expose as much or as little as they want, from email and calendar data to bank details; access is scoped to individual chats, and members can see when someone's personal information is being requested. Wabi says it is SOC 2 compliant and runs entirely in the cloud, with most user data stored in the U.S.

The launch is also a pivot. When TechCrunch covered Wabi's first release last year, alongside a $20 million pre-seed round, it looked more like a social platform for building and sharing mini apps. Kuyda frames the shift as evolution: people want software made for them, but may not want to spend time building it themselves.

There is no business model yet. Wabi is free for now, and Kuyda said the company is in talks to raise more funds — a common sequencing in consumer AI, where distribution comes first and infrastructure costs arrive immediately.

The open questions are whether a multi-threaded, multi-player design actually fixes the quality decay of long conversations, and whether users will trust an agent with inbox, calendar and bank access once permission boundaries are visible to everyone in the chat.