Thank you to @nytimes for yesterday’s piece on Dazzle (@itsdazzleai) and for recognizing our team’s hard work. While it’s still early days, the user engagement and response so far have been terrific. I wanted to share more context behind one anecdote in the article—my daughter Marielle’s birthday slide. To be clear, I didn’t use Dazzle because I was overwhelmed by making a slide. I used it because summing up your child in a few bullets feels laughably limiting. And, since our team had just gotten an early prototype running, I wanted to see what deeper threads it might find. So, when Marielle and her twin sister Sylvana’s 4th-grade teachers emailed asking for a slide celebrating each girl for their December 10th birthday, I pasted the email into Dazzle to test its limits. I fully expected it to struggle, since most facial recognition can’t even tell my identical twins apart. Instead, Dazzle blew me away. Not only did it distinguish Marielle from Sylvana, it surfaced forgotten kindergarten projects (like the “snake denomizer” and the “sandcat playground”) and connected them to who Marielle is today. I combined Dazzle’s insights with a few of my own and showed the slide to @zackbogue. He was floored and asked how I remembered those details. I fessed up that the Dazzle prototype had found them, and we both marveled at what AI can unlock from our camera rolls. (Pictures attached of Dazzle’s response, plus the slide I made with its insights and my own.) The foundational paper on transformers is titled “Attention Is All You Need.” What makes photos + AI so powerful is that our camera rolls are a decades-long record of what we pay attention to. By analyzing that visual history, an AI agent can connect dots we didn’t even realize we were documenting. A birthday slide isn’t a common case, but here are a few that have stunned Dazzle users recently: - Personalized learning: Dazzle offered custom spelling lists and grammar concepts based on photos of a child’s schoolwork. - Home inventory: Given a book club reading list, Dazzle found a book the user already had and noted two others by the same authors, entirely from the backgrounds of their photos. - Design intuition: Dazzle noticed a new sofa in a user’s living room photo and recommended artwork to hang above it. - Wardrobe suggestions: For an esoteric costume party, Dazzle recommended what the user already had in their closet and could wear, based on the clothes in their past pictures. We take 5 billion photos a day, often wondering if we’ll ever look at them again. Now, AI can turn those snapshots into deep, personal context.