I was Facebook's youngest engineer at 17. I left Meta at a moment when AI's lead can change every few months.

• Michael Sayman taught himself to code at 13 after his family was evicted during the recession.
• The former Meta executive says AI is shrinking the gap between one builder and a full team.
• Sayman left Meta's Superintelligence Labs for Whop, betting this is another App Store moment.

This as-told-to essay is based on a conversation with Michael Sayman, 29, a former Meta product executive who left the company's Superintelligence Labs to become President of Product Ecosystems at Whop, a New York-based creator commerce startup. Sayman, who joined Facebook at 17 as its youngest-ever software engineer, is also the author of "App Kid," a memoir about growing up as the child of Peruvian and Bolivian immigrants. The following has been edited for length and clarity.

People think of Silicon Valley as a place you go to because you dreamed of it. That's not my story.

After the 2008 recession, my parents and I got evicted. That forced me, at 13, to figure out how to make money online. I taught myself to code from YouTube tutorials. I built a word game called 4 Snaps. It hit number one on the App Store and made enough money to help keep my family afloat. The possibility of being able to do that, to build something from nothing and monetize it as a kid with no connections and no resources, that's what eventually put me on the path to Facebook.

Zuckerberg flew me out to Facebook's headquarters in Menlo Park when I was 17 for a one-on-one meeting on campus. He wanted to learn how I'd built my top-charting social apps in high school. That was our first meeting, and he ended up offering me a job. I became what I think was Facebook's youngest software engineer at the time. People would get me bottles of wine as a joke because I was underage. It felt more like a playground than a company.

The first thing anyone did at Facebook was boot camp. You pick a team, and you get to work. But I didn't do that. I had a slide deck with my read on where the product was headed and what I thought the company should be building. I presented it to my boot camp mentor. He brought in his manager. His manager brought in his manager. Eventually, I was in a room with Chris Cox (Meta's chief product officer), Kang-Xing Jin (Meta's former head of health), and Julie Zhuo (Meta's former vice president of product design), among others.

They greenlit me to start a new team, focused on the emergent, ephemeral nature of sharing that was coming up from apps like Snapchat and Musical.ly. Within a few months of joining, I had monthly product reviews with Zuckerberg.

Had I been older, I don't think I would've just taken it upon myself to do any of that. I was an engineer who was supposed to pick a team and shut up. I just didn't know that yet.

Those first four years were incredibly formative. But what I learned, more than anything, was my reference point: how different or similar Facebook was to everywhere else. That only became clear later, when I went to Google, then left to found my own startup, SocialAI, and eventually returned to Meta.

When Meta brought my SocialAI team on board in late 2024, and I joined Superintelligence Labs, the company I came back to was not the one I had left. It's so much larger now, and because of that, the smallest changes have the biggest impacts at a scale that is genuinely hard for people to grasp. Watching the AI race from inside Meta, I kept thinking: this is almost like the company is seeing its younger self in the rearview mirror. All these AI startups are operating with the energy and speed that early Facebook had, while Meta itself can no longer operate that way.

There's also a key difference. What Facebook was building in those early years was a competition of network effects. You were building a moat. Right now in AI, there is no clear moat. Every couple of months, there's a different company in the lead.

What Meta does still have, though, is something no one else has: the social layer. As AI agents start acting on your behalf in the world — finding information, making transactions, interacting with other people's agents — the question of verification becomes everything. Who are you talking to? That's where Meta's network becomes uniquely valuable again. That was the angle that brought me back, and what I spent my time exploring with Nat Friedman (co-lead of Meta Superintelligence Labs) and Daniel Gross (Meta's vice president of product) at Meta Superintelligence Labs.

While I was there, I also built the Meta AI blue ring — the visual embodiment that appears when you interact with Meta AI across every app, on iOS, Android, and web. I built that largely by myself. The number of people who would've taken even a few years ago is completely different from today. That's the shift. AI isn't just changing what we build — it's changing who can build it, and how fast.

That's also what made this the right moment to leave and join Whop, a creator commerce startup based in New York.

I'm 29. In Silicon Valley terms, I'm practically ancient. And I've felt for a long time this itch to take everything I've learned over the past 15 years and go make something with it. I just never wanted to force the timing. But right now feels like the App Store moment of 2008, a window where a small team, with the right tools, has the kind of leverage that used to require thousands of engineers. I didn't want to miss it.

At a company like Meta, you're always choosing which creator ecosystem to build for — Instagram, Facebook, or WhatsApp. At Whop, I can think about building across all of them. That's where I want to be: helping people build and monetize their own thing, the way I had to figure out how to do at 13 with no playbook.

I used to think that inside these big companies, there was some secret key that they had it all figured out. After 15 years, I can tell you that's not true. We're all just people trying to figure it out.

The difference now is that the tools to try are more accessible than they've ever been. That's the bet.

Have a tip? Contact Pranav Dixit via email at pranavdixit@protonmail.com or Signal at 1-408-905-9124. Use a personal email address and a nonwork device; here's our guide to sharing information securely.