RECOGNITION
PHYSICS INSTITUTE

Essays / Mathematics

Is Math Invented or Discovered?

Jonathan Washburn · 2 November 2026 · 19-minute read

Watch the film of this post (13 minutes), or read on below.

Your coffee shop gives you a loyalty card with ten squares. Buy a coffee, get a stamp. Fill the card, get a free coffee.

Nobody ever explained how the card works. Nobody had to. The barista stamps square one, then square two, then square three. The tenth stamp comes after the ninth. You can't stamp your way back to a blank card. And two cards with different numbers of stamps are still different after each gets one more.

People designed the card: the ten squares, the free coffee, the rubber stamp. Nobody designed the order of the stamps. Nobody signed anything. There was no committee, and no vote on whether square nine comes after square eight. Every card in every coffee shop in every country follows that order, whatever else its designers chose.

your coffee shop's loyalty card 1 2 3 4 5 6 7 8 9 10 full card: free coffee RULES OF COUNTING sign here: designed by people:ten squares, a free coffeenot designed by anyone:the ninth stamp comes after the eighth

So where do the rules of counting come from? Did somebody make them up, or were they there all along?

The short answer

Discovered. Counting is the record of telling two things apart, again and again, and the rules of counting are facts about that record. The record is reality's permanent record of every event: nothing that happens is ever erased, and the past still exists. The next section gives the reasons. Here's the whole answer in six steps. The rest of this post shows how we know each one.

  1. Counting is a record. Every time reality tells two things apart, the record gains one fresh entry. That record is the counting numbers: a start, then one more, then one more, with no last count.
  2. The rules come with the record. The start is never "one more" of anything. Two different counts never step to the same next count. Every count is reached from the start. And 1 + 1 = 2 checks with no extra principle at all, meaning no added assumption.
  3. There is exactly one way to count. Anything with a start and a step that follows those three rules is the record again, lined up with it in exactly one way: any two ways of counting by the three rules line up count for count, and only the names and symbols differ.
  4. Debts and slices come free. A negative number is a pair of counts, what's paid and what's owed. A fraction is a pair of counts with a rule for when two pairs are the same. No new ingredient is needed to get either.
  5. Every principle past the fractions has a price. Each one is a power to know something about an endless list without reading all of it. The record doesn't have that power for free, so a proof that uses one has to assume it: that's the price. Each has a name, and the prices we've named are proved to be different from one another.
  6. The smooth number line is bought. No list holds every endless yes-or-no pattern, so no list holds the line. "Is this number exactly zero?" costs one price, and "is it smaller?" costs a bigger one. The numbers the record holds are the ones a program computes: the record holds a number when one finite rule pins it down as closely as you ask, and such a rule is a program.

Why this matters: it settles what math is. Math is a fact about reality, not a game people agreed to play: a game's rules are agreed, but the rules of counting are facts about the record itself, and anything that counts by them is the record again, so only the symbols are left to choose. Counting isn't a language we laid over the world afterward. It's how the world's own record is tallied: each count is how many entries the record holds, which is why it describes the world. And it draws a dividing line you can read, with everything up to fractions free and every principle past it priced: the smooth number line, the one math class hands you as if it were a given, is on the paid side.

Where the argument starts

This post stands on its own. It starts from a few facts, each with its reason. Part 1 of the series The Universe Writes in Pen sets them out at greater length.

The smallest event is one thing told apart from another, a single yes or no. You're doing it right now: every perception and every measurement tells one thing from another. Take a sip of coffee, wait ten minutes, take another: it's colder. To know that, something has to have kept a note about the earlier sip. That gives three plain facts:

  1. Things can be told apart. Hot isn't cold. This happened, and then that happened.
  2. Comparing two things needs information about both. Colder than what? You need the earlier sip and the later one. The least information there can be is one yes or no.
  3. Checking a note is a new event. Remembering the earlier sip doesn't change it. It adds a new moment after it.

Part 1 also teaches a fourth fact, separate from those three: the past still exists. Reality is a record of every event: each one leaves a complete trace, and the full record of what happened is part of reality now. The record is permanent: nothing in it is ever erased.

Physics gives a reason: its basic laws never destroy information. In quantum mechanics this is unitarity: the present fixes the past exactly, and running the laws backward recovers every earlier state. That's why physicists treat the black hole information problem as a real problem: losing information would break the laws. Unitarity preserves information about the past. It doesn't settle which outcome gets recorded next: in quantum physics the state evolves exactly, but the answer a measurement records is not fixed in advance.

The record keeps every entry for a reason of its own: it only grows. Each new entry is added with every earlier entry kept, and writing a new entry never alters an earlier one.1 That's a separate matter from what an entry is. Each entry is a comparison, and a comparison needs information about both sides, so every entry is written against what came before.

Step 1: Counting is a record

Counting is the record of one act: telling two things apart. Every time reality does it again, the record gains one fresh entry. That growing record is the counting numbers.2

Here's how we know. The counting numbers come straight out of the three plain facts, and the third is the step from one count to the next. The record holds every event, and checking a note is an event. So look back at the record and you've done something new, and that new act is one more entry. Look back at that, and there's one more. The record only ever grows by one fresh entry at a time, and an old entry never changes.

At any moment the record holds finitely many entries: some definite number, however large. There is no last count: whatever the record holds, one more look adds one more.

Start at the very beginning, before any entry. That's a blank record. Tell two things apart and there's one entry. Check what you wrote and there are two. Check again and there are three. A record's count is how many entries it holds, and "one more" is the only move there is.

I'll draw the record as a loyalty card, a drawing of the real thing: every stamp is one entry, each card below is one count, and each arrow is one fresh stamp.

the record, drawn as loyalty cards each card is one count: how many stamps it holds 0 one more 1 one more 2 one more 3 ... blank: the start each arrow adds one fresh stamp. an old stamp never changes.

Next, and why: textbooks give counting a list of rules that make numbers behave. Are those rules in the record too?

Step 2: The rules come with the record

Three rules of counting hold for the record, and each one is proved:

  1. The start is never "one more" of anything.
  2. Two different counts never step to the same next count.
  3. Every count is reached from the start.

And 1 + 1 = 2 checks with no extra principle at all.

Here's how we know. All three are visible on the card.

Rule 1. The blank card isn't a card with a stamp added. "One more" always lands on a card with at least one stamp, so no card ever steps back to blank.

Rule 2. Add a stamp to a card with 3 stamps and to a card with 5, and you get 4 and 6. Different cards stay different. Turn it around: if two cards match after one more stamp each, peel that last stamp off both and the cards underneath match too.

Rule 3. Every card you're holding got its stamps one at a time, starting from blank. No card turns up from somewhere else.

1. the start is never one more of anything 2. two different counts never step to the same next count 3. every count is reached from the start 0 3 4 one more never lands on blank 3 4 5 6 different in, different out 0 1 2 3 ... every card is on the stairs

Are we using counting to check counting? We are. The claim isn't that counting is built from nothing. Checking anything takes reasoning, and ordinary reasoning already counts its own steps. So the rules show up twice: once in the reasoning you check with, and once as facts about the record. The second is what this post is about: the record obeys all three rules, and you can check each one on the card.

As Step 3 shows, nothing obeys all three except copies of the record. Textbooks list these three among the Peano axioms, the standard rules of counting, written down by Giuseppe Peano in 1889.3 There they're rules you're asked to accept. For the record they're facts.

Now 1 + 1 = 2. Adding works by two moves, for any counts n and m. Adding 0 leaves a count alone: n + 0 = n. And adding "one more than m" means adding m, then taking one more: n + (one more than m) = one more than (n + m). Here's 1 + 1:

Step Why
1 + 1 the question
= 1 + (one more than 0) the second 1 is "one more than 0"
= one more than (1 + 0) the adding rule
= one more than 1 adding 0 leaves 1 alone
= 2 one more than 1 is 2

That's the whole check. It never says "this is either true or false" about something it can't look at. It never searches an endless list. It uses no hidden power. In Step 5 you'll see why that matters.

Next, and why: is this one way of counting, or the only one? If somebody else builds a counter, do they get our numbers or their own?

Step 3: There is exactly one way to count

Take any system with a start and a step: a place to begin, and a way to move on by one. There is exactly one way to send the counts into it that respects the start and the step. Count 0 goes to the start, and each "one more" moves you one step on in the system. Call that the map. If the system also obeys the three rules from Step 2, the map is a perfect copy. That means the system is the record again, lined up with it in exactly one way, and only the names and symbols differ.

Here's how we know, starting with a clock. A 12-hour clock has a start, 12 o'clock, and a step, one hour later. Send the counts into it:

  • Where does count 0 go? To the start: 12 o'clock.
  • Where does count 1 go? One step on from where 0 went: 1 o'clock.
  • Count 2? One step on from that: 2 o'clock.

Every move is forced, so there's exactly one way. Keep going and count 12 lands back on 12 o'clock, so count 13 lands on 1 o'clock.

But the clock isn't a copy of the record. Counts 0 and 12 land on the same spot, because the clock comes back around. It breaks Rule 1: the start, 12 o'clock, is one more than 11 o'clock.

Now an elevator that stops at floor 4. The start is floor 1. The step is up one floor, and at floor 4 the button does nothing. Floor 3 steps to floor 4. Floor 4 steps to floor 4. Two different floors step to the same next floor, which breaks Rule 2. Not a copy either.

And a stack of plates. The start is no plates at all, and the step is to add one. Send count n to a stack of n plates. Every count gets its own height. No stack steps back to empty. No two stacks merge. Every stack is built from empty one plate at a time.

All three rules hold, so the plates are a perfect copy, and the map is forced: exactly one way. The same proof works for any system that obeys the three rules, and it uses no extra principle, only those three rules: Rules 1 and 2 keep different counts on different spots, and Rule 3 leaves no spot unreached.

a stack of plates a 12-hour clock an elevator that stops at 4 0 1 2 3 start: no plates step: add a plate the map is a perfect copy all three rules hold 12 3 6 9 0 12 both land here start: 12 o'clock step: one hour later gets the map, but not a copy it comes back around: breaks rule 1 4 3 2 1 start start: floor 1 step: up one floor gets the map, but not a copy floors 3 and 4 both step to 4: breaks rule 2

So the test is the three rules. Count in any notation you like: 3, III or ///. If your counting obeys the three rules, it lines up with ours in exactly one way. If another civilization counts by the three rules, its numbers are ours in other clothes. Different symbols, different words, same count.

The clothes are invented. The count underneath isn't. There is only one.

Next, and why: counting numbers can't say "I owe you a coffee" or "half a muffin". Do we have to add something new to get those?

Step 4: Debts and slices come free

A negative number is a pair of counts: what's paid and what's owed. A fraction is a pair of counts too, with a rule for when two pairs are the same. No new ingredient is needed to get either, and there is only one way to do each.

Here's how we know. Start with debts. You and a friend keep trading coffees. Keep two counts: coffees you've paid for, and coffees you owe. I'll draw each pair on a note. The notes are a drawing of the real pair of counts.

(5 paid, 3 owed) leaves you 2 ahead. (4 paid, 2 owed) also leaves you 2 ahead. Are they the same number? Here's a rule that never subtracts: add each note's paid count to the other note's owed count. The first note's paid and the second note's owed make 5 + 2 = 7. The second note's paid and the first note's owed make 4 + 3 = 7. They balance, so the two pairs are the same number. Only counts, only adding.

A pair like this is a signed count. When the owed count is ahead, it's a negative number. (3 paid, 5 owed) balances against (0 paid, 2 owed), because 3 + 2 = 5 = 0 + 5. Both are the number we write as -2.

Now slices. A fraction is a pair of counts: how many slices you have (the top), and how many slices make a whole (the bottom). The bottom is never 0. One slice of a muffin cut in 2 and two slices of a muffin cut in 4 are the same amount. The same kind of rule says why: multiply the top of each by the bottom of the other. 1 × 4 is 4. 2 × 2 is 4. They balance, so 1/2 and 2/4 are the same fraction.

debts: pairs of counts 5 paid 3 owed 4 paid 2 owed paid of one + owed of the other 5 + 2 4 + 3 7 = 7: level, so the same number: +2 slices: pairs of counts 1 slice of 2 2 slices of 4 = 1 × 4 = 4 2 × 2 = 4 top of each × bottom of the other they balance, so 1/2 and 2/4 are the same fraction

And there is only one way to do each. Take any number system that can add, subtract and multiply, where adding 1 over and over never wraps around to zero, unlike a clock. It contains our signed counts in exactly one way. If it can also divide, it contains our fractions in exactly one way.

Next, and why: what's past this? Every new principle past this point is a power over endless lists, and each one has a price.

Step 5: Where the free part ends

Everything so far is free: none of it needed an extra principle. Every principle past it has a name and a price, and the prices we've named are different from one another.

Here's how we know, starting with the plainest endless list there is: a drummer's pattern, a beat or a rest at every step, forever.

an endless pattern: at every step, a beat or a rest beat rest ... 1,000 rests so far ? step 1,001? 1. Is the pattern silent forever? 2. If not, when does the first beat come? if a beat comes, listening will catch it. if none ever comes, no amount of listening says so.

Two questions about any pattern. Is it silent forever? And if it isn't, when does the first beat come?

Say you've heard a thousand rests in a row. Is the next one a rest? You can't know. Nothing stops a beat from landing at step 1,001. If a beat is ever played, listening will catch it eventually. If none ever is, no amount of listening ever tells you so.

So to answer the first question for every possible pattern, the record needs a power it doesn't have: seeing a whole endless pattern at once. The record holds finitely many entries at any moment, never an endless number, so it can check any stretch it has but not an endless list all at once.

A power like that is an extra principle: a power to know something about an endless list without reading all of it. A proof that uses one has to assume it, and that assumption is its price. Each extra principle gives the record a power like this one, and the powers come in different sizes.

Every proof about the record comes with a receipt, the list of the extra principles it leaned on. The receipt for 1 + 1 = 2 is blank. The receipt is read straight off the proof's steps, so a proof can't use a principle without it showing.

The power to answer the first question for every pattern is the first price. Call it the zero test, because a pattern that is silent forever makes the number 0 (Step 6 shows how). To answer both questions, "silent forever, or here's the first beat," takes a bigger power: the search.4 Anyone who has the search has the zero test, because the search's answer already answers "silent forever?" Having the zero test doesn't give you the search; the possible worlds below show it.

At the top of the price list sits the one everybody treats as free: every statement is either true or false. It contains the search, and so the zero test, and everything else on the list. Here's why it contains the search. Say it about the statement "this pattern has a beat somewhere." It's true or false, and that answer is exactly the search's: false is "silent forever", true is "a beat is out there", and listening finds the first one. In the record it's the most expensive item on the list.

Here's the price list drawn as a menu board, a drawing of the real list.

THE PRICE LIST counting numbers FREE 1 + 1 = 2 FREE debts FREE slices FREE past the fractions, every extra principle has a price is this number exactly zero? THE ZERO TEST is it smaller than that one? THE SEARCH every statement is true or false ALL OF THE ABOVE, AND MORE RECEIPT 1 + 1 = 2 owed: nothing

These prices are exact, and they're different. For every pair of the prices we've named, we've built a possible world where one holds and the other fails: a consistent mathematical world, which logicians call a model, built to have one principle and lack the other. A proof of the missing principle from the one that holds would work in that world too, where the missing principle fails, so no such proof exists. No price can stand in for another.5

Next, and why: most points on the smooth number line take an endless string of digits to write down. What does the smooth line cost?

Step 6: The smooth line is bought

The smooth number line is bought, not given. Counts, signed counts and fractions can each be listed one by one. The smooth line can't. Asking "is this number exactly zero?" of every number costs the zero test. Asking "is it smaller?" costs the search. And the numbers the record holds are the ones a program computes.

Here's how we know, starting with lists. Line up the counts: 0, 1, 2, 3, and so on. Line up the signed counts: 0, 1, -1, 2, -2, and so on, and every one gets a place. There's a rule that gives every fraction a place too. Each of them can be listed, one by one.

The smooth line can't, and the endless patterns show why. Suppose someone writes a list of every endless pattern: pattern 1, pattern 2, pattern 3, and on down the page. Here's a pattern that isn't on it. At beat 1, play the opposite of what pattern 1 plays at beat 1. At beat 2, play the opposite of what pattern 2 plays at beat 2. And so on, down the diagonal.

beat rest a list claimed to hold every endless pattern beat 1 2 3 4 5 6 ... pattern 1 pattern 2 pattern 3 pattern 4 pattern 5 pattern 6 new pattern ... opposite of the circled beat, every time. it differs from pattern n at beat n, so it is not on the list.

The new pattern differs from pattern 1 at beat 1, from pattern 2 at beat 2, from pattern n at beat n. It isn't on the list. Whatever list anyone writes, this move hands them a pattern they missed. So no list holds every endless pattern. This is Cantor's diagonal argument.6 It needs no extra principle. It's in the free part.

Now turn patterns into numbers. Write a 1 for every beat and a 0 for every rest, and read the pattern as the digits after a decimal point, like 0.0110100.... Two different patterns differ at some beat, so they make two different numbers between 0 and 1. So a list of every point between 0 and 1 would contain a number for every endless pattern. Pick those numbers out, in order, and you'd have a list of every endless pattern: the very list nobody can make.

This is also where the zero test gets its name. Make a number from a pattern. It is exactly 0 when the pattern is silent forever, and only then. So "is this number exactly zero?" is the very same question as "is the pattern silent forever?" Being able to answer it for every number is the zero test. Being able to answer "is this number smaller than that one?" for every pair is the search.

Fractions don't cost anything here: a fraction is exactly zero when its top is 0, and you can see that at a glance. It's only the endless numbers that cost.

So what does the record hold? A number the record can hold is one it can pin down as closely as you ask, all with one rule. Ask for ten digits, you get ten digits. Ask for a thousand, you get a thousand, each answer a finite piece of the record. (Any one batch of digits can just be copied out. The point is that one rule answers every request.)

That rule is a finite set of instructions that the record carries out one step at a time, reading entries and writing new ones, which is what a program is. Rules like that are exactly the programs: every such rule is a program, and every program is such a rule. √2 has one. So do π and the golden ratio.

Now count the programs. A program is a finite text, so programs can be listed. The points on the smooth line can't be listed, as the diagonal argument showed, so there are more points on the line than there are programs. Nearly every point on the line has no program at all: it's a number no finite rule can ever pin down.

Here is the exact claim. Take "real" in the record's own sense. A number is real when the record can hold it, pinned down by one rule as closely as you ask. In that sense, a number is real exactly when a program computes it. That part is proved. The one step still being worked out is the last, from the record's sense of "real" to the physical world's: that what exists physically is exactly what the record can hold, no more and no less.

Adding, subtracting and multiplying the numbers the record holds is free, and so is dividing by a number you can show is at least some fixed distance from zero (say, 0.001). What costs is deciding: is it exactly zero? Is it smaller? The smooth line as a finished thing, with every point on it and those tests built in, is the purchase. Now you can read the receipt.

So, invented or discovered?

Notation is invented. The count is discovered. The exact tests and the finished smooth line are a purchase, with the price tag showing.

This is what we call actual mathematics: arithmetic built from one act, telling two things apart, with every further principle named and priced.

Further reading: The World Underneath the Labels sets actual mathematics beside actual physics, meaning, ethics and economics.


  1. That writing a new entry never alters an earlier one is Theorem 2.14, "Posting is an isometry that keeps its records," in "Recognition Boundary Theory: How One Ledger Holds Many Perspectives," the technical manuscript behind this series, from our institute: a preliminary manuscript (28 October 2026, 402 pages), not peer reviewed, at recognitionphysics.org/publications/. ↩

  2. The construction of the counting numbers from distinction, with their three rules, is in Jonathan Washburn and Milan Lj. Zlatanović, "The δ-calculus: From distinction to arithmetic," Advances in Pure Mathematics 16(9), 705-739 (2026), and in Jonathan Washburn, "The δ Foundation: Distinction, Arithmetic, and the Limits of Abstraction," fifth edition, version 5.0 (Recognition Physics Research Institute, November 2026; a technical book manuscript, not peer reviewed), at recognitionphysics.org/publications/the-delta-foundation/. The book also covers the one-and-only count, signed numbers and fractions. ↩

  3. Giuseppe Peano, Arithmetices principia, nova methodo exposita (Turin: Fratelli Bocca, 1889). ↩

  4. In constructive mathematics the search is the limited principle of omniscience (LPO), named by Errett Bishop in Foundations of Constructive Analysis (New York: McGraw-Hill, 1967). The zero test is its weak form (WLPO). ↩

  5. The receipts, the prices and their separations, Cantor's argument, which numbers one program pins down to every precision, and the prices of the exact tests on the number line are in "The δ Foundation: Distinction, Arithmetic, and the Limits of Abstraction," at recognitionphysics.org/publications/the-delta-foundation/. The possible worlds are models of intuitionistic second-order arithmetic: three are built from topological spaces and one from a chain of truth values (the book's Theorem 7.38). ↩

  6. Georg Cantor, "Ueber eine elementare Frage der Mannigfaltigkeitslehre," Jahresbericht der Deutschen Mathematiker-Vereinigung 1 (1890-91), 75-78. ↩