Manifold: 495 days betting on the ideas that won't let me sleep

I don’t know whether 495 days made me better calibrated or just more sceptical. The two look the same from the inside.

Manifold is a play-money prediction market where every question about the future can become a bet. My streak sits at 495 days. I open it most mornings before work — slightly embarrassing to write, because I’m a public attorney in Rondônia and my mornings are supposed to start with petitions, not prediction markets. But there’s something Manifold does that legal briefs won’t allow: assign a percentage to uncertainty and then be held to it. In a petition I have to assert. In Manifold I can say “63%” and mean it precisely.

Reading back the markets I created, I found something I hadn’t noticed while creating them: they cluster around the moments when I was most anxious to be right about something.

Turning a question into a market forces you to pick a resolution condition, a deadline, a definition. You discover, mid-question, that you weren’t asking “will AI replace audiobook narrators?” — you were asking “what counts as replacement?” The market forces the answer. That is cognitive hygiene I didn’t know I needed.

Calibration, or just scepticism?#

Looking at the trade history, I see soft corrections that, in prose, I would probably have rationalized as “I always thought that.” I also see markets that resolved against me slowly, over months, and in the end only confirmed I was wrong without teaching me where. Calibration doesn’t always come with a lesson.

Manifold’s collective signal, even with virtual currency, often anticipates moves the press only notices weeks later. I believe that. I also suspect I’m cherry-picking the cases where it worked — which is exactly the kind of soft rationalization the trade history shows me making.

A good market is a market with a sharp resolution criterion. When I write a vague market, the vagueness comes back to bite: weeks of arguing about whether the event counts as the event. Probability is a language. Saying “I think so” and saying “62%” are different sentences, and the second commits more. A petition can hide behind “it is evident that.” A Manifold market cannot.

What I wasn’t expecting: 495 days of weekly corrections to have the texture of a slow argument with myself, not a series of lessons. I’m not more certain. I’m certain about different things. The previous me who opened Manifold on day one was calibrated the way someone who never misses a filing deadline is calibrated — disciplined, but not necessarily right.

That previous me wrote everything with slightly less certainty about what to be certain about. That might be progress.