Wabi: a shared AI assistant for plans, group chats and personal interfaces

Wabi is presented as a shared assistant for group activities. Its current homepage focuses on shared activities: planning a trip, arranging a stay, sorting dinner or turning a group joke into a game. That framing makes it relevant to situations where an answer needs to fit a group’s conversation, rather than remain in a separate one-person chat.

Shared tasks are the current pitch

The Wabi homepage describes a group chat that does more than talk, with the assistant returning an outcome in an appropriate interface. It also advertises home-screen widgets for iOS and Android and moving between a computer and a phone.

Those examples suggest a place to begin: a small group plan with clear constraints, such as dates, preferences and a shortlist. A proposed itinerary is different from an authorized reservation. The brief homepage does not establish the exact booking permissions, payment controls or service coverage for every example, so those details need checking in the actual product.

A broader personal-software vision

Wabi’s manifesto describes mini apps based on a person’s tastes, habits and context, supported by persistent memory. That provides background for its interest in interfaces that adapt to the task.

The manifesto is a statement of product direction. It should not be read as confirmation that every imagined mini app or personalization capability is already available in the current assistant. Keeping the current homepage and the broader vision separate makes the product easier to assess without overpromising.

How to evaluate the fit

Start with one shared activity and check whether everyone can understand the resulting plan, contribute changes and identify what still needs a decision. If widgets are useful to the group, inspect whether they expose the information people actually return to.

Wabi is most interesting where a conversation needs to become something usable: a plan, a shared interface or a next step. This guide summarizes the company’s public descriptions, rather than a completed account-level test of those workflows.