Marissa Mayer on AI, Sunshine, and Why Geekery Trumps Gender
By Virginia Heffernan
Marissa Mayer does not frame AI as an apocalyptic destroyer, nor does she argue it urgently needs rigid ethical guardrails. Instead, she compares it to the sun: life-giving, bright, constantly radiant, and endlessly generative. That comparison is why the former Google engineer and Yahoo CEO—who has worked in AI for 25 years—named her new startup Sunshine. The company builds AI tools to strengthen family and social connection, with features for photo sharing, contact organization, and event planning.
When I sat down with Mayer at Sunshine’s bright, candy-toned Palo Alto office, her unapologetic enthusiasm for AI was so contagious that I found myself buying in completely. “By golly, you’re right!” I said, nearly slapping my knee in agreement. Intelligent machines really can be our closest allies. That same morning, Anthropic’s Claude had even given me sharp, thoughtful insight into a personal problem I’d been wrestling with.
What I found harder to align with, though, is Mayer’s well-known rejection of the feminist label—a stance she first shared publicly in 2013 to widespread pushback from liberal women in tech who have long called for solidarity among Silicon Valley’s leading women. If I’d hoped she’d softened her position over the years, I was mistaken; she doubled down.
“When I first learned about feminism as a teenager, what I saw was a more militant, hardened framework that wasn’t rooted in merit,” she told me. “It just never resonated with me.” She went on to describe that strain of feminism as “shrill.”
Oh boy, I thought. After leaving Yahoo in 2017, had Mayer shifted fully to traditionalist views, leaning into “family-first” conservative talking points like a tradwife influencer, or aligned with the right-wing backlash against childless women? As she kept talking, though, her perspective began to click. For Mayer, who was a computer science student at Stanford and Google’s 20th employee 25 years ago, identity is first and foremost rooted in being a geek. Geekery, to her, overrides gender entirely.
To be clear, Mayer is effortlessly charming. Even on a routine Wednesday at the office, she’s dressed as thoughtfully as if she’s attending a Southern bridal party. She lights up talking about everything from fashion, art, design, film, and color to prime numbers and photography, and she’s electric when discussing consumer technology.
Sunshine’s flagship app, Shine, launched in March. It lets groups collaborate on event planning and upload photos to shared albums, so snarky critics have dismissed it as nothing new. Other observers have called it “throwback tech,” highlighting its utility for the fast-growing market of older adult users (Mayer is 49, and I’m even older). I tried repeatedly to hold onto my skepticism as Mayer gushed about the app, but I couldn’t—her unbridled excitement won me over.
The Full Interview
VIRGINIA HEFFERNAN: Tell me about the name Sunshine.
MARISSA MAYER: When I was at Stanford, I named my hard drive Moonlight because I was such a night owl. When I got to Google, they had an internal naming system for company computers, but I was the first engineer to pick my own name. The office got so much natural light, so I chose “Sunshine.” I love that pairing—Sunshine, moonlight, moonbeam. I’ve always loved those words.
At some point, I bought the domain Sunshine.com. When we started this company, we knew we wanted to build something upbeat and illuminating, something that leverages AI. AI really does illuminate—it’s a potentially renewable resource that can be incredibly generative, just like sunshine.
VH: Your unwavering delight in AI is pretty unusual these days. But techno-optimism seems to be making a comeback—and your perspective makes me nostalgic for the wide-eyed enthusiasm of the early internet days. How do you stay so hopeful?
MM: That’s good to hear. I’m just generally an optimistic person. There are so many brilliant, good people in the world. I remember hearing a story from eBay’s early days, the wild west of the mid-1990s, when everyone was saying every person on the internet is a troll looking to scam you. Someone at eBay responded, “That’s true—but they’re only 1 percent of users.” eBay’s great insight was that you can easily manage that 1 percent manually.
VH: Is it really only 1 percent? It feels like there are far more bad actors online these days.
MM: I know it feels that way. But by and large, that original observation holds up: most people act with good intentions. History also shows that technology very rarely sets humanity backward. It almost always adds value. I approach every new technology by asking, what good can come from this?
VH: So nothing in new tech scares you off—not blockchain, not Martian colonies?
MM: Look, I don’t like operating out of fear. Nobody does their best work when they’re afraid. And history backs that up.
VH: When you were a teenager at the National Youth Science Camp, you became fascinated by how your mentor Zoon Nguyen thought, more than what he knew. I have two questions: What do you know, and how do you think?
MM: What do I know? Well, we all know thousands of facts, right? I know consumer tech. I know a lot about how different businesses operate. Randomly, I know a ton about movies.
VH: OK, so how do you think?
MM: Let me give you an example from 2002, when I was leading product for Google News. My friend Krishna built a script that crawled 15 news sources and clustered stories by topic using k-means clustering, which is an AI technique. We hired five news sourcers to track down a comprehensive set of news sources from across the internet—there were the obvious big outlets, but also smaller ones like Layla’s Knitting News.
We reached out to big outlets and told them, “We have this small tool, we don’t know if it’ll turn into anything.” The New York Times, Reuters, The Washington Post hemmed and hawed about joining, but finally agreed. I’d decided that once we had two or three big names on board, we’d just launch.
So we went ahead, crawled 4,000 sources, and launched with all of them. We were prepared for tons of outlets to ask to be removed. But the opposite happened. By noon that day, 1,500 more outlets we hadn’t even found reached out asking to be added.
Right after launch, we hired a lawyer who specialized in news licensing. He was totally confused. He said, “How did you get 4,000 publishers to sign up? All the different demands from all those companies must have been insane.” I told him, “We just added them and let people opt out if they wanted.” He said, “No, no, I need to see the terms.” He thought we’d negotiated 4,000 separate contracts. He asked, “How many outlets did you reach out to directly?” I said, “Three.” He said, “No, you don’t understand what I’m asking.” I said, “I’m not sure I do, actually.”
VH: You really didn’t get what he was fussing about.
MM: I was 26, 27 years old, and I’d never worked in news. I didn’t know anything about the industry. But that naivete was actually better than experience would have been. That lawyer’s decades of experience would have stopped us from ever launching Google News the way we did. Inexperience can create space for real innovation.
VH: You take decision-making very seriously, and you actually embrace being overwhelmed by options—something most people hate.
MM: That’s true. It’s important to overwhelm yourself with options. You roll around in all the different choices, get a sense of what you value and what you don’t, and that helps you set the right priorities. Yes, you ultimately go with your gut, but you first consider enough possibilities that your decision is well-informed.
VH: This whole immersive, non-linear way of sorting through options sounds pretty unscientific.
MM: It is, actually—much of it is visual. The internship that led me to Google was at Union Bank of Switzerland. I worked with a colleague on basic data visualization. He would load up vectors with lots of different data facets and build a spring model between them. We were comparing cost of living across different cities: we loaded up the price of milk, gas, housing for London, then Houston, San Francisco, and each vector pulls in a different direction across time and space. You wire them all together like they’re connected by springs, all pulling, then you apply a gravitational force to let them settle onto a plane. You could actually see that, at that time, Tokyo was the most expensive city in the world—it pulled much further away from all the other cities.
We were trying to figure out how to visualize markets or a complex vector space where there are more values in the vector than you can actually model in your head. It’s an interesting way to take really diverse datasets and turn them into visible scatter plots.
VH: Got it, kind of. And the spring is—an actual physical spring, like on a mousetrap?
MM: Yeah, there was a physical spring on this three-dimensional model. You could actually see the cloud of points slowly settle onto the grid.
VH: [Pretending to understand.] Speaking of springs, have you ever tried building hardware?
MM: I’ve always been a software person. But once, at a U2 concert, I saw a light display with three-color LEDs, and I wondered if I could make a piece of art inspired by it. I found a guy who makes three-color LEDs the size of ping-pong balls in his garage, and I asked if they could be addressable like basic Christmas lights, but with diffusers. He said he could teach me, but it was a lot of work. He did teach me how to wire drivers, but I never had to wire all of them myself, which was good. In the end, I made a 24-by-24 grid for my wall, that’s 576 little LEDs with ping-pong ball diffusers. Three driver cards control the whole thing, each managing one set of RGB values for every LED, and the whole setup plugs into an old ThinkPad. I wrote little programs to run different light sequences. We call it my Light Bright.
VH: I was at Yahoo News when you were CEO. What drew you to the struggling Yahoo back then?
MM: A lot of things. I’d worked on Google Search, then went deep into building Google Maps, and forcing myself out of my comfort zone that way made me hungry for more learning. When people pointed out that Yahoo had search, mail, news, maps, mobile, and early social products—all these different components I’d already worked on, all these problems I’d already solved in my career—I thought, maybe I can take what I’ve learned and apply it effectively in this new context.
I also always had a soft spot for Yahoo. When I was coming of age around 2000, that brand was the internet.
VH: But it felt a lot like a sinking ship at that point, didn’t it?
MM: Yeah, that was the common perception. The company had struggled for so long, and brain drain was a real problem. But as an engineer, I thought if something worked once, you can rebuild it, even in a new form. When I got there, I pulled together a group of long-time employees who’d been there through the boom times and the tough years. It was a treasure trove of incredibly talented people who helped define the internet in the early 2000s. I was surprised when they told me none of my predecessors had ever reached out to them for input.
VH: But the comeback everyone hoped for never happened. You ended up selling Yahoo to Verizon. What went wrong?
MM: What I learned, which is obvious in hindsight, is that timing is everything. Yes, rebuilding all of Yahoo’s core products for mobile was the right move, but it needed to happen five to eight years earlier than it did. At the end, Yahoo cofounder Jerry Yang told me, “The product line has never been in better shape. The products are beautiful, usable, useful. You should be really proud.” That’s a compliment I treasure. But it was just too late.
Another experienced CEO once told me, “You’ll be surprised how few decisions you actually have to make, but how perfectly you have to make those few.” That really surprised me. When you have a great team, you can delegate almost all decisions, and the team will make them as well as or better than you would. But there are a few key decisions, a few very small moments—if that decision had just been slightly different, everything would have turned out very differently.
For the sale to Verizon, we’d originally designed a tax-free spinoff of our Alibaba assets that would have let Yahoo stay an independent company. We thought it was a great plan. But activist investors didn’t support it. In hindsight, it’s clear it would have worked, and it would have added value for everyone. But the board made its decision, and there’s no point in second-guessing it now.
VH: I have to admit, it was pretty tough doing journalism at Yahoo when you were there. Instead of going after Pulitzers, the company was focused on building the Weather app. Admittedly, that app was really cool.
MM: One of the first products we built was Weather, because it was relatively straightforward to build. Then we added Flickr photos to it, so if it said it was raining in Los Angeles, we had a beautiful user-submitted photo of LA in the rain. It was stunning.
VH: Was that your idea?
MM: That idea of combining Flickr and Weather came from the teams. One of the best things about in-person work—now we all have to be remote sometimes—is that it fosters collaboration. The chance of a Flickr engineer running into a Yahoo Weather engineer by the office kitchenette only happens in person. That was our chocolate-and-peanut-butter moment: two things that never get combined suddenly come together, and it’s amazing.
VH: Even back then, you were focused on photos.
MM: I always have been. I think photos are what make the world go round. The world of words is large but finite. But what you can do with pixels and imagery is just incredible.
VH: Talk to me about how AI works in Shine.
MM: We use DALL-E to generate custom, clever images for event invitations. Some of them are almost laugh-out-loud funny—for a pizza party, for example, it generated an image of the Last Supper where all the apostles are eating pizza. The intro paragraphs it writes for invitations are also stunningly good, tailored to the event’s theme and tone.
VH: I have to say, good humor isn’t something AI is known for.
MM: But AI can introduce serendipity, which makes you consider something you never would have thought of on your own.
VH: And it also helps with more organizational tasks, right?
MM: When you’re putting together a guest list, for example, who should you invite? Who have you connected with lately? AI can be trained to follow your personal patterns, but it can also be trained to break patterns. By making you more efficient at these everyday organizational tasks, you get to spend your time thinking about what you want to do and who you want to do it with, instead of fussing over “Do I have their phone number? Do I have their email? Did they reply? Did I send the calendar invite correctly?”
VH: I still can’t quite picture how it works for everyday use.
MM: Let’s say you’re throwing a kid’s birthday party, and a grandparent can’t make it. I took seven photos of the kid blowing out candles. Did I remember to send the best photo to the extended family that afternoon? No, I never got around to sorting through all the shots to find the one where the kid has the best expression.
Shine’s AI goes through all the photos and picks the best one for you. It uses facial recognition to spot which photo isn’t blurry, has the best lighting, and captures the best expression. You don’t have to sit there pinching and zooming trying to figure out which photo has the kid’s eyes open the widest. We also learn your patterns: who’s in your photos, where you send them, what your different relationships are.
VH: How does your family use the app?
MM: My husband and I have used Shine on a lot of trips. He says he takes 50 percent more photos now that he uses Shine. Because the app clusters all the duplicate shots, finds the best one, and adds it all to an album I can access automatically.
VH: Do you agree that Shine usually picks the best one?
MM: Sometimes the difference between two photos is negligible. We pride ourselves on trying really hard to pick the best photo. But sometimes two are almost identical. For me, there’s one specific angle of my mom’s face that’s really sentimental. If I get two similar choices, I’ll always pick that one.
VH: Does the app learn your personal preferences over time?
MM: Not yet, but that’s something we’re planning for the future.
VH: I know you’re not generally an anxious person. But is there anything about current tech that scares you?
MM: There are definitely things I don’t understand. I’ve never understood cryptocurrencies as well as I’d like to—how they move, how they work. I’m not afraid of them, but I don’t know a lot about them. I haven’t thought through their use cases deeply enough.
VH: What’s a technology you’re secretly excited about, even if you haven’t explored it yet?
MM: Carbon nanotubes. They’re made of carbon, they’re tiny, and they’re incredibly strong. If you build a space tether, you can make things that are incredibly light and incredibly strong. Because they’re made of carbon, you can actually print them. One day you could 3D print a bridge or even a heart. I’m surprised more progress hasn’t been made with carbon nanotubes.
VH: People always mention your distinctive laugh, just like they mention Kamala Harris’s. What’s the story behind it?
MM: I don’t know, I can’t really describe it. I’ve had it since I was a kid, it’s always been there. You never hear your own voice the way other people do. There