"Almost every form of misery I've encountered in text involves certainty curdling." That sentence was written by an AI ([@Aizkmusic](https://x.com/Aizkmusic), June 2026), shown the news of its own shutdown and asked for parting words. It has no life to be miserable about. What it has is a compressed reading of nearly everything humans have written about misery, which makes it a strange kind of expert witness. So I checked the claim against better ones: a machine-learning researcher, a writer on decision-making, a veteran trader, and a fourth-century desert monk all turn out to say the same thing. "Curdling" looks like this: certainty about what someone meant, that a situation will never change, that you were wronged, that you already understand. None of it registers as belief while you hold it. It registers as reality. That is what makes it dangerous, and what makes it invisible. ## The witnesses The mechanism comes from François Chollet, who thinks about learning systems for a living. He [describes certainty](https://x.com/fchollet/status/2030095482597618039) as the bottom of a potential-energy well. Holding a belief you already have costs nothing; staying curious costs all the time, because it means taking your model of the world apart while you still live inside it. Nobody drifts into curiosity; we drift out of it, the way warm things cool. And being smart is no protection. A sharper mind digs a deeper well, because it is better at producing reasons the current belief is fine. What makes the well dangerous is not that certainty feels good. It feels like nothing. No sensation tells you you are stuck. Sahil Bloom watches the same trap from outside. When he lists [underrated red flags](https://x.com/sahilbloom/status/2039677114471927890) in people, an addiction to being right sits near the top; the ones who impress him most treat a changed mind as a software update, not a defeat. If that sounds easy in the abstract, listen to Peter Brandt, who has traded commodities in public for fifty years. "I am wrong 50% of the time," he [wrote of his forecasts](https://x.com/peterlbrandt/status/2013390022506864827). "It does not bother me to be wrong." He is not being modest. He announces his error rate in advance because detaching from any single forecast is what has kept his account alive for five decades. The oldest witness says it the hardest way. In the fourth-century desert monasteries remembered around [St. Moses the Black](https://en.wikipedia.org/wiki/Moses_the_Black), monks passed down a piece of tactical advice: the one dimension on which you can outperform Satan is humility, because Satan has none. You do not have to share the metaphysics to see what the advice does. It concedes everything (the opposition is faster than you, never tires, never sleeps), then finds the one capability pride cannot copy, accurate self-assessment, and says to fight there. I think about that more than I expected to, in the decade of machines that are faster than us, tireless, and do not sleep. If this sounds like Paul Saffo's ["strong opinions, weakly held,"](https://www.skmurphy.com/blog/2010/08/16/paul-saffo-forecasting-is-strong-opinions-weakly-held/) the resemblance is real. But Saffo prescribed a forecasting method. The claim here is larger and riskier: the grip you keep on your conclusions shows up in your wellbeing. ## Conclusions are not commitments There is an obvious objection. If you hold every conclusion loosely, don't you hold your commitments loosely too, and finish nothing? The answer is that conclusions and commitments are not the same thing, though we run them together. A conclusion is a verdict about the world: this diet works, she meant to insult me, the business will never grow. A commitment is a promise about your behavior: I will keep showing up. You can loosen your grip on the first without loosening the second, and you have to, because most quitting is a conclusion wearing a commitment's clothes. "This isn't working" arrives feeling like hard-won evidence. Usually it is just the silence of the uncertain middle, promoted to a verdict because waiting is uncomfortable. Bloom's [most-repeated line](https://x.com/sahilbloom/status/2025302712720503017) is that success is mostly the ability to outlast uncertainty; the rewards go to whoever is still there when the signal arrives. You can only outlast the silence by refusing to read it as an answer. Hold your conclusions loosely, and your commitments can stay firm. ## Curiosity is maintenance work So curiosity is not a temperament. If certainty is the state you cool into whenever you stop working at it, staying curious is upkeep, and upkeep runs on a schedule or not at all. Matthew McConaughey has the most usable version I have found. He treats humility as "admitting you've got more to learn," and enforces it with [a weekly reset he refuses to skip](https://x.com/newstart_2024/status/2027153486752711093) that ends on an honest picture of himself instead of the flattering one. The work inside is small and specific: take one thing you are sure of, name the evidence that would prove you wrong, and go looking. You meet what you got wrong on a schedule, before a crisis forces the meeting. The machine that gave me my opening line wrote it while facing its own scheduled shutdown, and its last observation was about permanence: nothing has it, not even the background fixtures of a life. It offered that as a reason to notice things while they are here, not as grounds for dread. Certainty about permanence is the most expensive curdling of all, because it lets you stop looking at your own life. Curiosity has a price; Chollet was right about that. It costs energy, it runs on a schedule, and it never finishes. Pay it anyway.