Dazzle mines your camera roll while Instinct bets big on travel

Two routes into the personal agent market

Two stories from TechCrunch this week map out where consumer AI assistants are heading, and they point in noticeably different directions. Former Yahoo CEO Marissa Mayer has taken the wraps off Dazzle, a personal assistant that derives everything it knows about you from a single source: the camera roll on your phone. Meanwhile Noah Shinn, founder of the viral invite-only agent Instinct, says his platform is approaching a billion dollars in annual transactions, with travel making up more than half of them.

Dazzle's bet on photos

According to TechCrunch reporter Marina Temkin, Dazzle raised an $8 million seed round last December and is only now being shown publicly. What separates it from the crowd of assistants that arrived over the past month — including Meta's Muse and Instinct — is where it draws context from. Rather than email, calendars or shopping histories, Dazzle analyzes the photos stored on a device to infer hobbies, style and food preferences, travel history and the people you spend time with.

The product splits into two functions. For immediate tasks, it scans recent photos to extract actionable details, such as adding an event to your calendar from a photographed flyer or finding a repair person after noticing a broken garage door. For longer-horizon planning, it mines the whole library to generate personalized suggestions for vacations, gifts and outings.

In Temkin's testing, the assistant recommended Mediterranean destinations that matched her past trips to Spain and Greece, and even surfaced Sicily from a visit four years earlier. It also had gaps: it suggested roller skates as a birthday gift for her daughter without registering that the child already knew how to skate. Mayer told TechCrunch that suggestions drawn from her own photos included escape room venues her family had never heard of.

Dazzle is Mayer's second attempt at a photo-centric product. Sunshine, her previous startup, launched an AI photo-sharing app called Shine in 2024 that drew criticism for its dated design and was eventually shut down, though Mayer says it produced useful intellectual property.

Privacy is central to the pitch. Mayer argues that users uneasy about granting an assistant access to email and messages may find a photo library a more comfortable trade, and the company says Dazzle discards personal information its systems flag as sensitive.

Instinct leans into transactions

Per TechCrunch's Ivan Mehta, Shinn told investor Patrick O'Shaughnessy on a podcast that travel bookings represent over half of all transactions on Instinct, and that the platform's annual transaction volume is nearing $1 billion — though he did not explain how that figure is calculated. He also claimed the platform is growing roughly 10% day over day, with transaction volume climbing at a similar pace.

Shinn framed the appeal as urgency: rather than browsing aggregated listings the way a travel site presents them, a user could simply tell the agent they need to be in New York that night. He went further with restaurants, suggesting an agent could poll reservation systems every few seconds, and floated a model that would give preferential treatment to people booking for special occasions. Critics pushed back, noting that agents could just as easily fabricate an anniversary to grab a table, and that many restaurants actively value their regular local customers.

The company behind the hype remains small: Instinct has 14 employees and recently closed $1 billion in Series C funding at a $10 billion valuation, with Sequoia Capital, Benchmark Capital and Coatue participating. It competes with Meta's Muse and Khosla-backed Wajo.

Why it matters

The rush of personal agents arriving within weeks of each other suggests the category has entered its first real differentiation phase. Dazzle is wagering that intimacy — an assistant that genuinely knows you — comes from an underused data source, while Instinct is demonstrating that agents can move substantial money through a narrow, high-frequency use case like travel. Each carries distinct risks: Dazzle's inference is only as good as its photo analysis, as its blind spots show, while Instinct's ambition to reshape reservation systems raises questions about what happens when agents start optimizing against the businesses they transact with. How privacy expectations settle — photo libraries on one side, payment and booking data on the other — may well decide which model consumers actually trust.