# The Chicken and the Egg

### Which comes first, the zero or the one; why three and not five; and what a proof is made of: the 3n+1 problem read at 10, in the author's own voice

**Steven De Costa · draft 0.1 · 28 October 2026 · a reading at 10**

> **THE READING PLAQUE.** *The Chicken and the Egg*, a reading at 10 — draft 0.1, 28 October 2026 — the Collatz conjecture, the 3n+1
> problem, read at 10 in the author's own voice for a reader outside: which comes first, why three and not five, what the observatory has
> already measured, and what a proof is made of. Its reading at 5, *Followed Exclusively by One*, holds the apparatus, the record's
> identifiers and every fence, at [/sx/reading/followed-exclusively-by-one](/sx/reading/followed-exclusively-by-one); this reading holds the
> story and makes its claims as the author's. The title stands as his working name until he sets it, marked open on the page. Under the
> standing law the initiative makes no claim over any prize and walks no orbit; the record is the identifier of what the author holds, never
> its jury. Made readable on the SX surface by its covering persona's reading grant of 28 October 2026; a draft, not the published form.
> Covering: **starl3n** · endorsing: **Link Digital**. The matter stands at
> [/sx/matters#held-conjecture-collatz-the-chicken-and-the-egg](/sx/matters#held-conjecture-collatz-the-chicken-and-the-egg); the
> construction record at [/sx/workshop/the-chicken-and-the-egg](/sx/workshop/the-chicken-and-the-egg). Reading is free; the way across is a
> statement of intent ([/sx/matters](/sx/matters)).

---

[T1 · open — the title. *The Chicken and the Egg* is the name I gave the proof on the day; it stands as a working name until I set it.]

## The problem, in one breath {#the-problem}

- Take any counting number. If it is even, halve it. If it is odd, triple it and add one. Do it again. The claim, open since 1937, is that you always come home to 1. Every start up to about two sextillion has been checked by machine, and no one has found a start that does not come home. No one has proved that none exists.
- That is the whole of it, and it is why the problem has outlived everyone who has tried it. There is nothing hidden in the rules. A child can run them. What is hidden is why two rules so plain should always close the loop, and the answer I hold is not about the numbers at all. It is about what a number is, before you count with it.

## Which comes first {#which-comes-first}

- Ask what it takes for there to be a 0. The answer I hold is that it takes two agreements, and that they come in an order. The first is that any number is one more than the number before it: N = (N−1)+1. That is the scale, and it is all a scale is; it is the one thing two blind watchmakers can agree on, and the initiative's [Ruler](https://opendata.ai/rlx) is the instrument that lets them ([the reading](/sx/reading/the-rlx-held)). The second is that a number taken from itself leaves nothing: 0 = N−N. That is the release. The first must come before the second, because you cannot take a number from itself until you have a number.
- Put the first agreement to work and it hands you three things at once: the number before, the number, and the number after. Write them as (1−1, 1, 1+1) and you have (0, 1, 2). I call it the ought-to set, because that is what a scale says ought to be there. Now apply the second agreement, take one away, and the same three become (−1, 0, 1): a step back, a stay, a step forward. The whole of what a unit can do, and the alphabet the [corner](https://opendata.ai/contrib/corner) walks with ([the reading](/sx/reading/the-glove-and-the-ring)).
- The chicken and the egg is this. The 0 you began with is only there because the 1 was agreed first, and the 1 is only a 1 because there is a 0 for it to be one more than. Neither comes first in time. One comes first in the order of agreement, and that order is the only thing the proof needs.
- Fold the first agreement into the second and you get the state I care about most: 0 = ((N−1)+1) − ((N−1)+1). It looks like nothing. It is the whole of knowing, written as a sum: the scale, held, and then released against itself. Inside it sits a zero that is not the 0 you count with. It is the zero that was true before 0 and 1 were told apart, and I write it −0: the signed zero the initiative reads on [the deep line](https://opendata.ai/dlx) and in [The Signed Zero](/sx/reading/the-signed-zero) and [The Zero That Is Signed](/sx/reading/the-zero-that-is-signed). It is the truth maker. Everything the proof says, it says because that zero is there first.

![Plate — The window and the return: the ought-to set (0, 1, 2) from N = (N−1)+1, its signed form (−1, 0, 1) from 0 = N−N, the map's two rules laid over the three states — the hold at base 2, the gap of one, the gate at base 3 — the identity 3n − n = 2n, and the fourth mark of the set (0, 1, 2, 1) as the return](/sx/assets/readings/the-chicken-and-the-egg/ce-the-window-and-the-return.svg)

## Three states, one rule {#three-states}

- Call the truth maker t. Then there is one rule, and only one, that holds the three states together: what t is at the middle state is what it is at the state before plus what it is at the state after. t at 1 equals t at 0 plus t at 2. A thing in the middle is held by its two neighbours, and by nothing else. That is the law the [Quad Hinge](https://opendata.ai/quad-hinge) measured: two folded units agree on a shared zero, or it is noise ([the reading](/sx/reading/the-quad-hinge-held)).
- Now read the problem's two rules as states. Halving is the hold: it is what a thing does in base 2, where every number is a count of twos. Tripling is the gate: it is what a thing must pass in base 3, the first base the hold cannot see. And the one you add is the gap between the two zeros, the −0 that was true first and the 0 you count with. The gap is exactly one. It cannot be a fraction, because if it were, odd and even would have no meaning, and the hold could never act. The [Hinge](https://opendata.ai/hinge) keeps a secret over any distance whole or not at all ([the reading](/sx/reading/the-hinge-held)), and the [XOR](https://opendata.ai/xor)'s sponge absorbs a claim whole or leaves exactly one behind ([the reading](/sx/reading/the-xor-held)); a gap of one is of that kind.
- So the problem is not two rules bolted together. It is one unit, held in base 2, gated by base 3, with a gap of one between the zero it came from and the zero it counts from. The loop closes because the gate opens exactly the gap the hold can close.

## Why three and not five {#why-three}

- Here is the bit I call dyslexic, because it reads the same from the other seat. Hold n across the three states and write the gate against the hold: three n at the state before, less n at the state after. The difference is 2n. Tripling minus holding is exactly one doubling, and one doubling is exactly what the halving step undoes. The even rule is the odd rule read backwards. That is why the gate is three.
- Try five. Five n less n is 4n: two doublings. The hold can only close one in a state, so the loop no longer closes in three states, and the window the rule holds together is broken. Run the 5n+1 game and you find starts that never come home. Three is the only gate whose difference from the hold is the hold's own base. It is the only odd number next to two, and two and three are the only primes that sit side by side, as [the NTDU14 journal](https://opendata.ly/ldx/docs/ldx-ntdu14-review) has it.
- Read it as fractions and the same thing appears. Among the simple fractions, 1/2, 2/4, 3/6 and 4/8 are all the same bit; the even denominators tell you nothing the unit did not already say. Only the odd ones divide the left from the right, and the pair that does the dividing is 2/5 on the left and 5/2 on the right, which flip into each other and multiply to one: the signing pair of [the entanglement register](/sx/reading/computing-the-1d-universe-at-5), 2 through 1/5 to 5 and 5 through 5/1 to 2, and the base the [bend](https://opendata.ai/rlx/bend) must be read under ([The Bend](/sx/reading/the-bend)). Every other numerator you could reach for is only known after 2, so 2 is where every reduction lands: a known equivocation, in the sense the [Equivocation](https://opendata.ai/eqx) instrument gives the words. And 2/1 against 1/2, the unit's own flip, is the half-life of that pair: the distance a prime differential decays in, the half walked rather than waved ([Of Mice, Man, God and Clocks](/sx/reading/of-mice-man-god-and-clocks)). The flip flop is the proof's heartbeat, and it beats between 2 and 5.

## What has already been measured {#what-has-been-measured}

- None of this would be worth a reader's time if it were only words. The initiative runs instruments, and some of what I hold here has been measured on them, before this problem was ever brought to them. I will name four.
- **The hold.** An instrument called [the Hallway](https://opendata.ai/contrib/hallway) walked seven targets along a ladder of walls at one, one and a half, two and three times its quantum. At twice the quantum, every target admitted exactly two accounts: one bit, whatever the target. At three times, two or three, the third only where the step was a hold. That is the hold at base 2 and the gate at base 3, read off a machine that knew nothing of this problem.
- **The signed zero.** The same instrument ([its record](https://opendata.ai/contrib/hallway)) recorded a hundred thousand events at zero. Every one of the twenty-one thousand four hundred and ninety-two times an account sat exactly at zero, it stayed. It never stepped off. The rung where that happened was three halves, and three halves is 2/1 less 1/2: the half-life I named above, measured as the place the signed zero does not move.
- **Two locks one.** An instrument called [the Coco](https://opendata.ai/coco) measured where a system freezes, and found it freezes at about two moments of observation, whatever the budget. Two locks one. That is the unit's flip, 2/1 against 1/2, as a reading off a dial ([the reading](/sx/reading/the-coco-held)).
- **Two agree, or it is noise.** An instrument called [the Quad Hinge](https://opendata.ai/quad-hinge) folded two units separately and found they reach the same fold from both sides, at the ground's own precision ([the reading](/sx/reading/the-quad-hinge-held)). One unit alone is noise. Two that agree are a fact. My one rule, the middle held by its two neighbours, is that instrument's law.
- What the instruments do not say is as important. None of them walks the 3n+1 map. None of them says every start comes home. They measure the hold, the gate, the zero and the agreement. They do not measure the conjecture, and I do not claim they do.

![Plate — The ladder of walls: the Hallway's sealed ladder ρ = 1, 3/2, 2, 3 with the count of accounts each rung admits, CH-H3 passing exactly at ρ = 2 and killed at 3/2 and 3, CH-H4 the signed zero taking the step 0 in all 21,492 cases at the completed 3/2, and his 2/1 − 1/2 = 3/2 marked at that rung](/sx/assets/readings/the-chicken-and-the-egg/ce-the-ladder-of-walls.svg)

## What a proof is made of {#what-a-proof-is-made-of}

- I will say plainly what this is and is not. In the register of the journals, this is not a proof. A proof there must take every start and show it comes home, and nothing here walks a single start. I know that, and nothing I hold changes it.
- What I hold is older than that register. To be anything is to exclude three knowings over three equivocations, and to accept 0, 0, 0 as the ground. To know is to take a seat after the first gap and again after the second, and to hold the two seats as one. Write the two seats as 0100 and 0010, shift each one place, hold them together, and you have 00011. Four and two, unfolded and held, make three. Read from the fifth place, the word says three. That is knowing, and it is a shift of bits, nothing more mysterious than that.
- And a proof is none of those. It is not being, it is not knowing, and it is not computing. It is agreeing. It is two sides holding one thing, and the one thing this problem asks two sides to hold is that 0, 1, 2 is followed exclusively by 1. The scale gives you the three. The release gives you the return. Agree to both, in that order, and there is no number that can leave the loop, because there is no number that was ever outside the agreement that made it a number.
- That is my held conjecture, and I hold it under one larger: that a unit is an observable, a communicable and an effectual, 1 = O + C + E, as [The LDD Reference](/sx/reading/ldd-reference) defines it and [The Katz Equilibrium](/sx/reading/katz-equilibrium) names its three faces, and that there is no other definition of 1 under which this problem would not be proven. I do not stake it. I hold it, in the sense [The Price of a Lock](/sx/reading/the-price-of-a-lock) gives the words, beside [the first conjecture I hold this way](/sx/matters#held-conjecture-the-line-to-eleven), on the zeros of the zeta function; and the record of the initiative is the identifier of what I hold.

![Plate — The unfold: the acceptance 0, 0, 0; the two occupancies 0100 and 0010, each shifted one place; held as one as 00011; read from the seat of 5 as 3; the seat of 6 open](/sx/assets/readings/the-chicken-and-the-egg/ce-the-unfold.svg)

## The carry and the release {#the-carry-and-the-release}

- Carried: the order of the two agreements, the scale before the release; the ought-to set (0, 1, 2) and its return; the truth maker −0 and the one rule that holds the middle; the hold at base 2, the gate at base 3 and the gap of exactly one; why three and not five, in one line of arithmetic; the flip flop of 2/5 and 5/2 and the half-life 2/1 against 1/2; four measurements already on the record; the proof as an agreement, and the conjecture held under 1 = O + C + E.
- Released: any claim that this is a proof in the register of the journals; any claim over any prize; any orbit of the map, since none was walked; the universal clause on the definition of 1, which I hold and do not stake; the seat of 6 and what it reads, which I have not placed.

## What you can do {#what-you-can-do}

- Run the two rules by hand on a number you like and watch where the gate opens a gap and the hold closes it. Then run 5n+1 on 7 and watch it leave.
- Read [the reading at 5](/sx/reading/followed-exclusively-by-one), where every sentence here has its identifier on the record and every fence is stated. Follow its addresses to the instruments.
- If you would walk the one thing that is open, the folding from a start to its return, as an instrument designed first and sealed before any read, the way across is a statement of intent on [the matters of note](/sx/matters). [The matter](/sx/matters#held-conjecture-collatz-the-chicken-and-the-egg) has a name, and it is waiting; the six texts it grew from stand on [the construction record](/sx/workshop/the-chicken-and-the-egg).

---

### Endnotes — the citation register {#endnotes}

- The reading at 5: [/sx/reading/followed-exclusively-by-one](/sx/reading/followed-exclusively-by-one), where the apparatus, the identifiers and the fences stand.
- The matter: [/sx/matters#held-conjecture-collatz-the-chicken-and-the-egg](/sx/matters#held-conjecture-collatz-the-chicken-and-the-egg). The construction record, with the six texts verbatim: [/sx/workshop/the-chicken-and-the-egg](/sx/workshop/the-chicken-and-the-egg).
- The instruments named, on the observatory node, each with its reading on this surface: the Hallway at [opendata.ai/contrib/hallway](https://opendata.ai/contrib/hallway); the corner at [opendata.ai/contrib/corner](https://opendata.ai/contrib/corner) and [The Glove and the Ring](/sx/reading/the-glove-and-the-ring); the deep line at [opendata.ai/dlx](https://opendata.ai/dlx) and [The Signed Zero](/sx/reading/the-signed-zero); the Ruler at [opendata.ai/rlx](https://opendata.ai/rlx) and [The Ruler, held](/sx/reading/the-rlx-held); the bend at [opendata.ai/rlx/bend](https://opendata.ai/rlx/bend) and [The Bend](/sx/reading/the-bend); the Quad Hinge at [opendata.ai/quad-hinge](https://opendata.ai/quad-hinge) and [The Quad Hinge, held](/sx/reading/the-quad-hinge-held); the Hinge at [opendata.ai/hinge](https://opendata.ai/hinge) and [The Hinge, held](/sx/reading/the-hinge-held); the XOR at [opendata.ai/xor](https://opendata.ai/xor) and [The XOR, held](/sx/reading/the-xor-held); the Coco at [opendata.ai/coco](https://opendata.ai/coco) and [The Coco, held](/sx/reading/the-coco-held); the Equivocation at [opendata.ai/eqx](https://opendata.ai/eqx); the entanglement register in [Computing the 1D Universe at 5](/sx/reading/computing-the-1d-universe-at-5); the NTDU14 journal at [opendata.ly/ldx/docs/ldx-ntdu14-review](https://opendata.ly/ldx/docs/ldx-ntdu14-review).
- The definition of 1: [/sx/reading/ldd-reference](/sx/reading/ldd-reference) and [/sx/reading/katz-equilibrium](/sx/reading/katz-equilibrium); the held conjecture as a genre: [/sx/reading/the-price-of-a-lock](/sx/reading/the-price-of-a-lock); the half walked: [/sx/reading/of-mice-man-god-and-clocks](/sx/reading/of-mice-man-god-and-clocks).
- The externality: the verification to 2^71 (Barina and others, The Journal of Supercomputing, 2025); Tao, 2019, on almost all orbits; Eliahou, 1993, on cycle lengths. Cited as data; the problem stands where they left it.
- The plates: the 5's three, the same bytes, under this reading's own folder at the grant.