Culture > People > Work
A senior leader at Meta shared a framework with me that I keep coming back to. Get the culture right first, and you'll attract great people. Great people do great work. The order matters.

For a long time I thought the opposite. Hire exceptional people and everything else follows. What I've seen over and over is that the culture outlasts the people. You can bring ten great leaders into a broken org. Most will leave, adapt, or burn out trying to change it. The culture wins.

So what actually is culture? It's the values on the website, but it's also the unspoken actions that are exhibited within an organization. Instagram has always been explicit about craft and design, and you feel that in how the product gets built and who thrives there. The problem is when the stated values and the lived reality diverge. Culture is ultimately what gets rewarded. Who gets promoted, what behavior gets tolerated, how conflict gets handled when no one senior is in the room.

The reason culture comes before people is that it shapes who joins, who thrives, and who stays. Anthropic attracts a different kind of person than Meta. Meta attracts a different kind than Stripe. That's not accidental — it's the culture selecting for itself over time. The talent profile of a company is mostly a lagging consequence of its culture, not the other way around.

Work is even further downstream. When a company is shipping great things, you're usually watching the output of cultural decisions made years earlier. And when things start to slip, the work is almost always the last indicator to move. Culture degrades first. People follow. Work is just where it finally becomes visible.

The place I find this framework most useful is inside large organizations. Meta isn't one culture. Instagram, WhatsApp, Reality Labs, and Meta Superintelligence each have different risk tolerances, different definitions of winning, different kinds of people who thrive there. When two of those orgs need to work together, they're not just aligning on a roadmap. They're navigating what are essentially two different companies. Most cross-functional failures get written off as execution problems. In my experience, they're usually culture mismatches that nobody named.