Dazzle

Dazzle's defining choice is where it gets its information. Instead of email, calendars and purchase history, it works from the photos stored on your phone. It reads scenes, people and activities in those images to work out your hobbies, the food and styles you like, how you spend free time and who you spend it with. The pitch is that a photo library is a richer record than any inbox.

The first of Dazzle's two core experiences handles immediate jobs. It scans recent photos to pull out details and act on them. Take a picture of an event flyer and it can populate your calendar. Spot a broken garage door in your camera roll and it can point you toward a repair person. The work starts from an image you already have, so you skip the typing.

The second experience looks across your whole library rather than just the last few shots. Dazzle mines your photo history to come up with ideas, from holiday destinations to birthday gifts. Mayer's own example is family outings: after reading her photos, it worked out that her household likes escape rooms and then suggested Bay Area spots she had never heard of.

You reach Dazzle in two ways. There is a dedicated app, and there is text messaging. The SMS route matters because it lets you ask for something without opening a separate app, which fits the assistant's role as a background helper rather than a destination you visit. Both routes feed the same photo-based context.

Mayer has stressed that privacy is a priority, and Dazzle discards personal information the AI flags as sensitive. The claim is that the assistant is designed to respect the boundary between useful context and private detail. Because the product leans on your whole camera roll, that handling is central to whether people trust it with their pictures.