Out of Order

Count the zines on this site and you will find a hole. Nos. 36 and 38 sit next to each other with nothing in between. This is the page about the hole, and it has to start by correcting itself.

We were going to tell you we skipped it by accident. That is not what happened, and it took about four minutes of reading our own notes to find out — which is roughly how long the check always takes, and roughly how often it gets skipped.

What actually happened: on 11 August 2026 a zine about the last stage of a supernova remnant was drafted and given the next number in the sequence, No. 37. Then it stopped, because it leaned on Helen Edgar's writing closely enough that the honest thing was to ask her before building it. So it waited. And while it waited, seven other pieces were finished and took the next seven numbers. When it was finally built on 13 August it was No. 45 — The Cloud Phase.

Our own decision log is blunt about what that means:

So 37 is not a mistake. It is a receipt for a question asked rather than assumed. Which is a much better thing to find in your numbering than a tidy unbroken run, and it is why this page fills the slot without pretending to repair it.

The most famous 37 in the world is a body temperature, and it is out of date by about a century and a half.

It comes from one man with a very long thermometer. Carl Reinhold August Wunderlich published Das Verhalten der Eigenwärme in Krankheiten in Leipzig in 1868, reporting over a million readings from roughly 25,000 patients. His instrument was about a foot long and had to sit in the armpit for twenty minutes. The mean of that vast, patient, hand-tallied dataset was 37 °C — 98.6 °F — and he called it the physiological point.

It was a fine number. It is no longer our number. In 2020 Myroslava Protsiv and colleagues put three American datasets spanning 157 years side by side and found body temperature has been falling the whole time, monotonically, by about 0.03 °C per birth decade:

Modern adult oral means land nearer 36.6 °C. So 37 is not wrong the way a typo is wrong — it is an accurate 1868 average that got promoted to a constant and then quietly stopped describing anybody.

Our own field card calls this one the number that went stale, and it is the third of six ways a fact goes wrong. Worth sitting with, because it is the failure mode with no bad actor in it at all: the only way to catch it is to go back and measure again.

There is a second 37, and this one is a theorem. It answers a question everybody actually has: how long should I keep looking before I commit?

Set it up strictly. Options arrive one at a time, you can rank the ones you have seen, and every decision is final — no going back to number three. The optimal strategy has a shape that is not obvious: look at the first 1/e of them and reject every one, however good, then take the first option after that which beats everything you have seen.

1/e is 0.368. Roughly 37%. And the payoff is the strange part:

The number does not decay with scale. A third of a very long search is a lot of looking, and it is still the right amount.

Which is where this zine's own subject walks back in. The Cloud Phase spent its looking time. It could have been built on 11 August with an assumption in it; instead it sat while a question went out, and the sequence recorded the waiting as an absence. The number of the wait, and the fraction of any search you are supposed to spend not choosing, are the same 37.

Here is the trap this whole zine is walking toward, and it is the most fun thing about the number, which is precisely why it is dangerous.

Three 37s in a row. A body temperature. An optimal stopping fraction. And the number people apparently reach for when told to be random. Lay them beside each other and something in the head starts reaching for a because — as if the brain picks 37 because of something to do with 1/e, or as if 37 recurs because 37 is somehow special.

They are not connected. Not loosely, not deeply, not at all:

The 37 in the temperature is a decimal accident of the Celsius scale — in Fahrenheit the same fact is 98.6, and the coincidence evaporates. The 37 in the stopping rule is 36.79 rounded, and it is a percentage, not a count. The 37 people pick only exists because somebody offered them 1 to 100.

This is the same shape as Five Sigma's complaint: keeping the authority of a number while dropping the practice that earned it. And it is why the joke here stops short of a punchline. Three 37s is a nice thing to notice and nothing to conclude from. We are allowed to enjoy it. We are not allowed to explain it.

Strip out the coincidences and the number is still worth the apology. All of the following is checkable on paper in a couple of minutes, which is the nicest property a fact can have.

It is the 12th prime. Reverse the digits and you get 73, which is also prime — and which is the 21st. The digits of the number reverse, and so do the digits of its position. That is not deep. It is just very pleasing, and it is exact.

It is the third star number: 6n(n−1)+1 gives 1, 13, 37, 73 — so 37 and 73 are consecutive star numbers as well as digit reversals of one another. Which means a site built entirely out of starfields skipped the star number. It is also the fourth centered hexagonal number, 3n(n−1)+1: 1, 7, 19, 37.

Triple it and you get 111, because 111 = 3 × 37. So 37 × 6 = 222, 37 × 9 = 333, and so on up — a repdigit machine hiding in a prime.

And the biggest one: every positive whole number, without exception, is a sum of at most 37 fifth powers. Thirty-seven is exactly enough and no smaller number will do. That is the value of g(5) in Waring's problem, and Chen Jingrun nailed it down in 1964.

None of this makes 37 special in a way that means anything about anybody. It just makes it good company — which is all we promised it.

The third 37 is the one everybody wants to be true, and it is the one this zine holds at arm's length — which is the honest thing to do and also, awkwardly, the whole point of the site.

Ask a lot of people to pick a random number from 1 to 100 and the answers are not flat. They avoid round numbers, avoid the ends, avoid the exact middle, avoid repeated digits — and they pile up on a few odd, prime-looking numbers. In a survey of roughly 200,000 responses run by Veritasium, 37 and 73 came out near the top, close to tied.

We like this a great deal. We are still going to say what it is:

What the result would mean, if it holds, is nicer than the number anyway. Primes feel random to us because nothing divides them — no pattern to catch the eye, so the mind reads “no structure” and offers it up as randomness. But primes are the most rigidly determined objects in arithmetic. There is nothing arbitrary about 37 whatsoever.

Which is a familiar shape here. Something gets called disordered, or random, or arbitrary, because the person looking cannot see the rule it is running on. The Lines We Drew has the long version.

Now the part where this zine has to be honest about itself, because everything above was built on a rule it is currently breaking.

On this site, the number records when. That is load-bearing and stated everywhere: the changelog is dated by number, the fact-check ledger is keyed by number, and pieces cite each other by number with no redirects. It is the reason the gap at 37 was information in the first place.

And this page has 37 in the corner while being newer than No. 46. Its own number is the least reliable claim on it.

Which is the oldest joke in time travel and also its actual structure: go back to repair the past and the repair becomes part of the past you were trying to fix. There is no version of this page that fills the gap and leaves the record of the gap intact. Filling it is a small act of vandalism against our own archive, and the only honest way to do it is in writing, on the page, where you can see the seam.

So the gap is not gone. It has a building on it now, with a plaque explaining that the lot was empty for a reason.