There is an assumption built into every founder's mental model of their own career, so deep it's never even examined: that the judgement problem gets easier as the company succeeds. More resources, better people, more experience, more pattern-matching in the tank. The early years were the hard ones: deciding everything, alone, with nothing. Surely altitude brings air.
It brings the opposite, for four reasons that compound each other, every one of them invisible from the inside.
Start with the obvious one: the stakes scale and the instrument doesn't. The decision muscle you had at seed is the same one making the calls at scale: same physiology, same capacity, same vulnerability to a bad week. But each rep now carries ten or a hundred times the load. A misjudged hire once cost you a salary; it now costs a division a year. Sport has no equivalent to this. No athlete's final gets objectively harder each season as a reward for winning the last one. Yours does. Success is a ratchet on the value of every future judgement, applied to a fixed piece of equipment.
Second, and less obvious: your feedback loop is dying. In the early days, decisions taught you. You chose, and reality graded the paper within weeks. The customer churned, the feature landed, the hire flamed out in a quarter. Fast, attributable consequences are the only conditions under which judgement actually improves. At scale, the consequences of your biggest calls take years to arrive, and they arrive laundered through a hundred intervening events (market shifts, execution layers, luck) so that even in hindsight nobody can honestly say which judgement produced which outcome. The result is quietly brutal: the learning mechanism that built your judgement switches off at precisely the altitude where errors become unaffordable. You are flying on instruments calibrated at sea level.
Third: the information itself degrades. Every layer of organisation between you and the ground is staffed by people with careers, and bad news climbs an org chart the way cold water climbs a hill. What reaches you has been summarised, softened, and framed by people who know exactly how it will land. None of this requires dishonesty, only ordinary self-preservation, multiplied by layers. The founder of a two-hundred-person company is making the largest decisions of his life on the most curated inputs of his life, and the curation is invisible by design.
Fourth, the cruellest: correction disappears. Success buys deference. The track record that got you here becomes the reason no one in the room presses the weak point in your argument. After all, you've been right before, and they were there to see it. The pushback that once stress-tested your thinking dries up person by person, promotion by promotion, until the only voice reliably interrogating your judgement is yours. Athletes never lose this; the coach's entire function is to be undeferential, and no gold medal exempts you from Tuesday's video review. Founders lose it on a schedule, and experience it as harmony.
Now add the four together. Bigger bets. Worse data. Slower learning. No correction. The top job, arrived at through good judgement, is the most hostile environment for human judgement anyone has ever constructed, and it is handed out as a prize. Every promotion toward it strips away another of the conditions under which deciding well is even possible. It's as if a sport existed where each victory raised the quality of your opponents while removing a member of your coaching staff.
You cannot fix most of the altitude problem. The deference is structural; the feedback lag is physics; the information filtering is what organisations are. Founders who understand the problem usually attack those fronts (skip-levels, kitchen cabinets, red teams) and those tools are useful, but they are patches on an environment that will keep degrading as the company grows, because degrading is what growth does to the founder's decision conditions.
Which leaves exactly one variable in the whole system that is fully yours, fully improvable, and gets more valuable as everything else gets worse: the state of the instrument doing the judging. When the inputs are curated, the feedback is gone, and the room agrees with everything, the margin between a good call and a bad one lives entirely inside you: in the capacity, on the day, of the one system nothing at altitude protects. Lower down the mountain, managing that system was an edge. Up here, it's the only handrail left.
The air doesn't get thicker. You get built for altitude, or you don't.