Conviction or Delusion

Every founder who was ever wrong felt exactly the way you feel right now.

That’s the sentence nobody wants to hear, so it’s the one this post is built around. Certainty is not evidence. The deluded founder and the visionary founder report identical internal states. They cite the same legends, keep the same rejection emails, feel the same fire. So “do I believe in this?” is a worthless question. Belief is table stakes in both populations. The question that separates them is: what, outside your own head, would change your mind?

I set this up in Everyone Said No First with a concession about survivorship bias and then moved on. This is the post that stays. Chesky kept his rejection emails and was vindicated. A hundred founders kept theirs and were simply wrong. Persistence appears in both groups at full strength, which means persistence is not the differentiator, even though a lot of founder advice treats it like one.

What separates them, as far as I can tell, is where they look for the verdict. Investors have opinions. Customers have behavior. Usage, retention, repeat payment, unprompted referrals. Behavior is the only data that doesn’t care how the meeting went. When the room says no and the usage says yes, keep going. When the room says no and the usage also says no, the room may be early to a conclusion you need to reach yourself.

The sharpest tool I know for this is stolen from science: falsifiability. A founder with conviction can name the evidence that would make them quit. Named in advance, specific, checkable. A founder in delusion cannot, because every negative signal gets reinterpreted on arrival: they don’t get it yet, the market isn’t ready, the messaging was off. A belief that can’t be falsified isn’t conviction. It’s faith wearing conviction’s clothes, and faith is fine, but you should know which one is running your company.

Now the part that complicates all of it, and I refuse to hide it: markets are late. Behavior data said no to plenty of things that needed years to become legible, and the “usage will tell you” rule applied strictly in year one would have killed some category-defining companies. Pivots wreck the clean binary too. Slack came out of a failed game. Instagram came out of a cluttered check-in app. Was that conviction or delusion? Neither. It was delusion that updated, which is a third category the title pretends doesn’t exist.

So the honest version of this post isn’t a verdict test you pass. It’s a practice. Early on, behavior data is thin and noisy at exactly the moment rejection is loudest, and the skill is holding the belief loosely enough to update and tightly enough to keep building through the noise. Those two motions feel opposite because they are. Nobody does this well. I want to be clear that I don’t do this well, and I’ve been at it a long time.

It rhymes with something I wrote about code: the move is separating the work from the self, so the idea can be wrong without you being wrong. Easier with a function than with the company you’ve bet years on.

And the close can’t be clean, because the truth isn’t. You can build the falsifiability test, weight behavior over opinion, hold the belief loosely, and still lose, because the market’s verdict arrives on its schedule, not yours. Conviction is a position you maintain under uncertainty, not a fact you verify in advance. The test never finishes running.