Eugenia Kuyda pivots Wabi toward multiplayer AI agents

Wabi 2.0 represents a departure from the company’s original vision as a social platform for building mini-apps. While initial feedback showed users appreciated personalized software, they often balked at the effort required to create it. Kuyda’s solution is an automated system: the agent now builds its own tools upon connecting to services like email, spinning up specific threads for tasks such as health tracking or family scheduling.

Kuyda positions the platform against competitors like Muse and Instinct, arguing that single-thread AI models inevitably degrade as they accumulate excessive context. By segmenting interactions into dedicated threads, Wabi aims to prevent the hallucinations and information overload common in endless, monolithic chat logs. The core innovation, however, is the multiplayer functionality. Rather than keeping AI interactions private, users can invite others into these threads, enabling shared access to calendars, music trackers, and relational tools. This structure allows participants to query the agent about group-specific logistics, such as checking a partner’s availability or confirming shared purchases, with security scoped to each individual chat. Wabi is currently free for all users as the company explores future monetization strategies.