AIBRIEF
The veteran investor is backing ex DeepMind engineer Shivani Poddar's Wajo, whose assistant Fo calls businesses, hires people, and pays on the user's behalf.
Personal AI agents have stopped being research demos. In the last quarter they have become a contested market, and Vinod Khosla's public bet on Wajo, a startup from former Google/DeepMind engineer Shivani Poddar, marks the moment the differentiator shifted from capability to data stewardship.
Wajo's assistant Fo lives in iMessage and WhatsApp and presents work as task cards on a web dashboard. It calls businesses to book or cancel services, can hire a human when the task requires it, issues a single-use virtual credit card so the user's real number never hits the checkout page, and keeps its own email inbox. When a website blocks the agent, Fo either nudges the user to act or opens a three-way chat with the business for verification.
Wajo claims 10x growth across 107 countries since a late-September launch, with no user count disclosed. The arXiv paper on the trust architecture has not been independently benchmarked. Khosla contrasted Wajo with Meta's Muse and OpenAI's upcoming Dot, arguing both have weaker data-handling contracts; funding was not disclosed, though Jeff Dean and Gokul Rajaram are also backers.
The market is now real and crowded. What is still unproven is which of these agents can hold a user's trust when something goes wrong on a real chargeback, not on a demo.