The conversation before the split

Noam Wasserman spent a career studying why promising startups die and put a number on the unglamorous truth: roughly 65% of high-potential startups fail because of people problems — cofounder conflict chief among them. Not the market. Not the model. The two people who started it, no longer able to work as one.

If you are somewhere on that road now, you have probably noticed the strange shape of the advice available to you. Your lawyer can structure a separation beautifully — vesting, buyouts, IP, the works — but can't tell you whether to have the conversation at all. Your investors have a portfolio-sized interest in the answer. Your other executives can't be told it's even a question. Your partner in the conflict is, obviously, disqualified. And the friends who love you will tell you what people who love you tell you.

So the equity gets advised, and everything that actually determines whether this goes well goes unadvised: the sequencing, the sentence you open with, what you owe each other after seven years, whether the resentment you're carrying is a signal about them or about the role you've outgrown, and the grief — because a cofounder split is a divorce, and the people who have been through one will tell you the settlement was the easy part.

Those people exist. Someone bought out their partner in 2021 and can tell you which of their fears materialised and which evaporated. Someone was bought out and can tell you what the initiating founder did that made it survivable — and the one thing that still stings years later. Someone chose to repair instead, and can tell you honestly whether that was wisdom or postponement. You don't know these people. Structurally, you can't know them: anyone close enough to know your situation is too close to advise on it. That is the exact gap a vouched, sealed panel exists to bridge.

They answer separately, without seeing each other's accounts, because on a question this emotionally loaded the last thing you need is a room that talks itself into consensus. You read them with the farthest view first. Then — only then — you have the conversation you've been rehearsing, carrying five maps of the terrain instead of one 3am monologue.

The conversation is probably coming either way; the only question is how prepared each of you is for it, and how much of the company's remaining life gets spent on the approach. Two weeks of sealed counsel is usually cheaper than another quarter of silence.

Pannels convenes a few people who have made your exact decision — sealed, so no view bends another. If you're facing this one now, pull up a chair — what you write is read by one person, and you can't say it wrong.