Wabi Launches a Shared AI Assistant Built Around Threads

Eugenia Kuyda launched Wabi for all users on October 19, 2026, a multiplayer personal AI assistant that works as a messenger app. TechCrunch

Wabi organizes help around multiple agentic threads, which are ongoing AI conversations that can remember context and take actions. Each thread handles a different task or long-term goal. Threads can be shared and work like group chats, with several people and the assistant working in the same conversation.

That structure returns to Kuyda's starting point. She presented a conversational chatbot called IO on stage at TechCrunch Disrupt in 2014. IO was designed as a lifestyle concierge to book restaurants and vacations and handle other tasks. It was the first product from her startup Luka. Luka later became Replika, the AI companion company.

Kuyda has since left her role as CEO of Replika and founded Wabi. Wabi raised a $20 million pre-seed round reported at its initial launch.

Wabi did not start as a messenger. It operated as a prompt-based app builder, where typed descriptions produced simple apps, before shifting to a personal AI agent that can create interfaces on demand. TechCrunch That change moved interface creation out of a separate builder step and into the conversation itself, with the assistant producing the screen or tool a task needs at that moment.

On its homepage, Wabi calls itself "The AI assistant you use together". Wabi The product runs on computers and phones, so users can start on one device and continue on another. It also offers widgets that can be added to iOS or Android home screens.

The broader context here is the shift from single-user, single-session assistants to persistent agents used by several people over time. Most assistants still keep context separate for each user and each chat. Wabi threads that last across long-term goals and include other people change that coordination. Information has to stay consistent across participants, devices, and time. Permissions, limits on what the AI remembers in each thread, and how long threads are kept become central design choices rather than extras.

In my view, the interesting technical bet is shared agency. A private thread is largely a problem of memory and planning. A shared thread is a coordination problem across people and devices. The assistant has to track who asked for what, who can see what, and which instructions still apply as the conversation moves on. Done well, that cuts the copy-and-paste work now common in group planning, household logistics, and small-team work. Done poorly, it creates confusion about who said what and who is in control.

Looking at what this means for builders, building screens on demand fits that model. A fixed app screen assumes one task and one user. A thread that spans tasks and participants needs temporary checklists, bookings, tables, and controls that appear, update, and disappear as work changes. Widgets and cross-device support point the same way. The assistant lives where people already pay attention, not inside a separate app they must remember to open.

Worth flagging is how familiar this cycle feels. Concierge bots in 2014 promised to handle bookings and errands through chat, but they relied on rigid scripts and human workers behind the scenes. Current models can understand intent, use software tools, and keep context over longer periods. The pitch has come full circle. What will decide Wabi is less fluent conversation than reliable follow-through across shared, long-lived threads.