Five people died this fortnight in a submersible whose builder had, on the record and with evident pride, dismissed the certification standards for such vessels as needless obstruction, an anachronism, safety-worship that stifled innovation. He said the rules were written for a timid industry and that at some point safety is just pure waste, and then he took his uncertified craft into the crushing dark and it did what the rules existed to prevent. I build things for a living, less dangerous things, and I want to write in defence of exactly the kind of rule he mocked, because his contempt for it is a specific and seductive and recurring error, and the deep sea merely punishes it faster and more visibly than most domains do.
There is a principle usually attached to Chesterton, about the fence across a road: the reformer who says he sees no use for the fence and will therefore tear it down should not be allowed to touch it until he can explain why it was put there, because the fence was almost certainly erected by someone to solve a problem, and the fact that the reformer cannot see the problem is more likely evidence of his ignorance than of the fence's uselessness. Most safety rules are Chesterton's fences, and the specific thing they fence off is a death that already happened. The certification standards for crewed submersibles are not arbitrary caution; they are a ledger, written over decades, of the precise ways the ocean has already killed people, each requirement a headstone, each specification the compressed and anonymised memory of a hull that imploded or a seal that failed or a material that fatigued, encoded into a rule so that the next builder does not have to rediscover, with his passengers, what the deep already taught at someone else's cost. To look at that body of rules and see only obstruction is to see a graveyard and mistake it for a parking lot.
I owe the counterargument its full strength, because I have felt its pull and it is not simply the vanity of one reckless man. Innovation does require breaking rules; every genuine advance looked reckless to the incumbents whose standards it violated; certification regimes really can ossify into gatekeeping that protects established players and strangles the new; and a civilisation too frightened to ever exceed its existing safety envelope would never have flown, never have sailed, never have gone to space, because all of those, at the outset, were statistically insane. This is true, and it is important, and it means the builder's instinct, that rules can calcify into cowardice, is not wrong in general. But there is a distinction he fatally missed, and it is the distinction between the rule you break because you have understood it and surpassed it, and the rule you break because you cannot be bothered to understand it and find your own ignorance more convincing than the accumulated evidence of the dead. The first is how progress happens. The second is how people die of things that were solved before they were born, and the tell that separates them is whether the rule-breaker can explain, in detail, why the rule was there and why his specific innovation genuinely renders it obsolete, or whether he simply finds the rule annoying and his own confidence sufficient.
Diane Vaughan, studying the Challenger disaster, gave the mechanism a name that every builder should carry: the normalisation of deviance. An organisation, or a person, departs from a safety standard, and nothing bad happens, because the margin the standard protected was large and the departure small, and so the departure comes to feel normal, and then a larger departure is taken from the new normal, and again nothing happens, and again it normalises, each successful violation eroding the perceived need for the rule, until the accumulated deviance finally exceeds the real, invisible margin, and the thing that had been fine every previous time is suddenly, catastrophically, not fine. The submersible had descended before and come back. That is not evidence that the rules were wrong; it is exactly the seduction Vaughan described, the deep rewarding the arrogance, again and again, teaching the builder that his contempt was wisdom, right up until the single dive on which the accumulated deviance met the true limit, which had been there the whole time, unmoved by how many times it had been flouted.
So I do not read this as a freak tragedy or as simple hubris to be clucked at from a distance. I read it as the clearest possible lesson in a discipline I try to practise: that the rules I did not write, the standards that predate me and seem to slow me down, are usually the encoded memory of failures I have been lucky enough not to have witnessed personally, and that my inability to see the reason for a constraint is the beginning of an inquiry, not the end of one. The deep sea is honest in a way most domains are not; it kills the normalisation of deviance quickly and without ambiguity. In the systems I build the deep is slower and the implosion quieter, a breach, a loss, a failure that unfolds over months, but the physics of arrogance is the same. The fence was built by someone. Before you tear it down, you had better be able to say who, and why, and what they lost teaching you the thing you are about to decide you already know.