In 2012, Evan Williams had already built two successful companies. He co-founded Blogger, sold it to Google, then co-founded Twitter and took it public. By any measure, he had earned the right to be confident. So when he launched Medium, he did it with a founder’s certainty: he knew how platforms worked, he knew how content spread, he knew what readers wanted.
Twelve years later, Medium is alive but not thriving. It has pivoted multiple times, alienated writers with shifting monetization rules, and never quite found a model that sticks. For a product built by someone with that track record, it’s a puzzling outcome. And it’s actually a pretty common one.
Serial founders fail at higher rates than the startup mythology suggests. The assumption is that experience compounds: you learn from your mistakes, you bring a network, you move faster because you’ve seen the patterns. Sometimes that’s true. But there’s a different set of problems that experience creates, and they tend to bite harder precisely because the founder doesn’t recognize them as problems at all.
The Setup: What Williams Knew
When Williams launched Medium, he was operating on pattern-matched intuition from Blogger and Twitter. He understood that content creation was hard for most people, that distribution was the real lever, and that the early community you attract shapes everything that follows.
These were real lessons, genuinely learned. And they weren’t wrong, exactly. The problem was that they were incomplete in ways he had no way to recognize. Blogger and Twitter had both succeeded in environments where the concept of a “platform for publishing” was genuinely new. Medium arrived into a world where WordPress had existed for nearly a decade, where Substack and Ghost would soon emerge, and where writers had developed strong opinions about owning their audience.
The patterns Williams recognized were patterns from a different competitive landscape. He was solving a 2004 problem in 2012.
What Happened
Medium launched as an invite-only, curated platform. It was beautiful and slow, and the deliberate exclusivity worked to attract a certain kind of thoughtful writer. Publications like The Ringer and Backchannel moved there. For a while, it looked like Williams had pulled it off again.
Then came the pivots. Medium tried being a publisher. Then a subscription platform for readers. Then a partner program that paid writers based on engagement. Then adjustments to that program that cut payments dramatically and drove away the mid-tier writers who had built real audiences there. Each change made sense in isolation as a response to a business problem. Together, they read as a company that never figured out who it was actually for.
The core issue wasn’t execution. It was that Williams had high conviction about certain product decisions, conviction built from prior wins, and that conviction made him slower to hear signals that contradicted his model. Writers were telling Medium early on that they were worried about owning their audience. The platform’s response was slow and unconvincing. A first-time founder with no prior wins would have been terrified by that feedback. Williams had enough confidence to treat it as a problem to be managed later.
Why the Second Time Is Structurally Different
There are a few specific dynamics that make the second startup harder, and they’re worth naming precisely because they masquerade as advantages.
Confidence in the wrong things. First-time founders are paranoid because they have to be. They question their assumptions constantly because they have no track record telling them their instincts are reliable. Second-time founders have evidence that their instincts have worked before. The problem is that instincts calibrated to one context don’t automatically transfer. The skills that made Blogger work were product intuition and timing. The skills Medium needed were marketplace dynamics and writer economics. Williams was confident in the former and underweighted the latter.
Network as a constraint. The second-time founder’s network is a genuine asset for fundraising and early credibility. It’s also a filter on the feedback you get. When you’ve already won, the people around you tend to assume you know what you’re doing. They’re less likely to push back hard on your assumptions. First-time founders often get brutal honest feedback because nobody has any particular reason to protect their feelings. The second time around, you’re surrounded by people who believe in you, which is nice, but it means the environment that would catch your wrong assumptions is compromised.
The speed trap. Serial founders move faster. This is usually described as an advantage, and in some situations it is. But speed means you commit earlier, you hire faster, you build more before you’ve validated the core assumption. The value of going slow at the start is that you stay exposed to disconfirming information for longer. A founder who spends eight months in awkward, manual, small-scale pilots learns things that a founder who spends two months on strategy and then raises a seed round will never learn. Seed rounds fund stories, and experienced founders tell better stories faster, which means they often raise money before they’ve genuinely stress-tested the idea.
The success model. Every founder’s second company gets compared, consciously or not, to their first. This creates a tendency to rebuild familiar structures even when the new problem doesn’t call for them. You know how to run a product team that ships weekly because that worked before. You know how to structure growth loops because you’ve built them. You build the scaffolding from memory. But the scaffolding was designed for a different building.
What We Can Learn
The founders who navigate the second startup well tend to share a specific trait: they treat their own experience as data to be questioned rather than rules to be applied. They’re willing to be beginners again in the domains where this new problem is actually novel.
Stewart Butterfield built Flickr as a side effect of a failed massively multiplayer game. He then built Slack as a side effect of a different failed game. The second success worked in part because Butterfield kept approaching the problem with genuine curiosity rather than certainty. He wasn’t trying to repeat Flickr. He was trying to understand the actual problem his team kept running into.
The practical implication is uncomfortable: the more successful your first startup was, the more deliberately you need to create conditions where you can hear bad news about the second one. That means talking to users who are frustrated, not just the enthusiastic early adopters. It means listening hard to the customers who leave. It means hiring people who will tell you you’re wrong, not people who are excited to work with someone who already has a W on the board.
Medium isn’t a failure in the dramatic sense. It’s still publishing, still paying some writers, still attracting some readers. But for the person who built Blogger and Twitter, it’s a lesson in what happens when you bring the right experience to the wrong problem and move too confidently to notice the difference.
The second startup is harder not because you know less. It’s harder because you know just enough to stop asking the questions that would save you.
