Personal AI assistant Dazzle, founded by former Yahoo CEO Marissa Mayer, raised $8 million in a seed funding round last December. Introduced on September 29, 2026, the product is based on the user’s photo archive on their phone rather than their email, calendar, and shopping history.
Dazzle aims to analyze photos to draw inferences about the user’s hobbies, areas of interest, food and clothing preferences, travels, their children’s interests, and who they spend their time with. The assistant, which can be used through the app or by messaging, has two main functions: carrying out immediate tasks, such as adding events to the calendar by extracting information from recently taken photos or finding a repair technician for a broken garage door seen in a photo; and examining the photo library to offer suggestions for vacations, activities, and birthday gifts.
Mayer said that after Dazzle scanned her own photos, it determined that her family liked escape rooms and suggested some venues in the San Francisco Bay Area that she had not previously known about. In one test, the assistant suggested Mediterranean destinations based on the user’s trips to Spain and Greece, but also listed Sicily, which the user had visited four years earlier. In another trial, it failed to remember that the user’s daughter already knew how to skate. Despite this, Dazzle offered personalized ideas such as a local ceramics studio and a bioluminescent kayaking tour in Tomales Bay.
Mayer’s previous venture, Sunshine, released the AI-powered photo-sharing tool Shine in 2024, but faced criticism, low usage, and closure. Mayer said the product nevertheless created “interesting intellectual property.”
According to Mayer, Dazzle could make sharing a photo archive more acceptable than granting access to sensitive data such as emails and messages. The company says it prioritizes privacy and deletes personal information that the AI flags as sensitive. Although Dazzle is not yet as comprehensive as other assistants, it points to a future in which AI could get to know the user rather than merely carry out tasks.
Why it matters
Dazzle’s distinguishing feature is that it seeks to generate personalization not from email or message content, but from the visual and social context conveyed by photographs. While this approach changes the question of which data is more acceptable to the user, it also shows that a photo archive can contain sensitive inferences about family relationships, habits, and past experiences. The company’s commitment to deleting sensitive information therefore creates a matter of trust independent of the product’s core functions; however, no details have been shared about how this is implemented. The errors in the trials reveal that drawing meaning from a large archive is not only a personalization issue but also a matter of accuracy. The closure of the previous photo-sharing venture, meanwhile, suggests that Mayer must once again prove user interest and long-term viability with the new product.