The Quarter, Three Ways — what I brought home from Number Theory Down Under

THE READING PLAQUE. The Quarter, Three Ways — what I brought home from Number Theory Down Under, draft 0.1 — made readable on the SX surface by its covering persona's reading grant; a draft, not the published form. The grant stands by the operator's word (2026-09-09), verbatim: "that gets a reading grant for the public sx surface." Covering: starl3n · endorsing: Link Digital (the standing pair). The essay is signed as the operator's own — Steven De Costa, in his own voice, never the 'we' of the Initiative — and was rebuilt from his earlier draft after a claim-by-claim audit against the public record; its construction record stands in the open workshop at /sx/workshop/the-quarter-three-ways. Every measured number links its sealed public record; every reading is marked as a reading; the sessions are credited by name for what each gave him, under the cares-are-mine law. The boundaries travel with the essay: no claim on any Millennium Prize problem, ever; no claim over the theorems of the discipline, over π, over the primes, or over physics. The source of record is ooi_network/docs/the-quarter-three-ways-draft-0-1.md (pinned sha256 LF b07f781c…); this file's sha256 is pinned in the reading register. Nothing here asks anything of any reader; reading records nothing about anyone. Reading is free; the way across is a statement of intent (/sx/matters).

Draft state — this block records the state of the draft; the essay begins after the rule. Draft 0.1 — 9 September 2026, venue trued the same day (the conference was organised by UNSW and held at its Canberra city campus; the first-0.1 mint, pin 385dcf02…, is superseded by this restamp, never erased). Author: Steven De Costa. Composed at the operator's word of 9 September 2026 ("this needs to be built as an md file … it should be a registered item being identified in the SX node workshop as it will be moving toward a public reading grant once approved, for the SX public surface") from his own earlier draft, The 25% Tax on Reality, and granted the reading surface the same day. Standing: a reading, stated never asserted; the experimental record it reads stands whole without it. A revision lands as 0.2; nothing is erased.

Last week I spent five days at Number Theory Down Under 14, organised by UNSW and held at its Canberra city campus, in my own town. I was not there as a number theorist. I run an open data company, and for the past year I have been running a research programme on the side called the Objective Observer Initiative, which walks small deterministic instruments along the number line and archives every hypothesis before it reads a result. I went to the conference to find out whether the things those instruments keep finding would survive a week in a room full of people who know the line far better than I do.

Three things came home with me: a defect, a quarter, and a line that is willed. Everything below is mine. Where a number is measured, I say so and link the record. Where it is a reading, I say that too. Nothing here claims anything over the theorems of the discipline, over π, or over the primes. The instruments are the subject. The mathematics is the ground they stand on.

The sessions that did the work

I owe the shape of this post to a handful of talks, and I want to say what each gave me rather than pretend the ideas arrived on their own. What I took from a talk is my own act. None of it is a claim about what the speaker meant, and none of it is a claim on their work.

Florian Breuer's talk on Ducci sequences and codes was the one I had been waiting for without knowing it. The Ducci map, where every cell of a ring becomes the absolute difference with its neighbour, is the mechanism inside every instrument I have built this year. I asked him in the room whether the three-cell case reads as a regression toward a holographic principle. He was generous with a question that was really mine to answer, and a stencil I drew during his hour became an instrument of its own by the end of the week (The T-Series Instrument).

Blair Butler's talk on estimating the ranks of elliptic curves refused a number I had been carrying. I will come to that below. It was the most useful refusal of the week.

Ade Irma Suriajaya's plenary on pair correlation is where a page of my notebook filled with the six nines of π, and that page became an instrument by the evening.

Felipe Gonçalves's plenary showed geometric and area-based ways of characterising a walk inside a hypercube, and I bought him lunch to talk about inference. Sam Chow's plenary on van der Waerden's conjecture and Snehinh Sen's talk on semiring modules set the frame for the rest: the first for what a random polynomial looks like from the discipline's side, the second for what arithmetic looks like when subtraction is taken away.

Persistence needs a defect

I arrived with a ponder, not a claim: that if reality sits on a plane confined between two boundary surfaces, and you fold that plane three times the way a Koch curve folds, the geometry of the resulting ring is isomorphic to running a Ducci sequence on a cycle whose length is not a power of two. I wanted to know what such a ring does.

So I built the ring and ran it. Three folds of a Koch curve closed into a snowflake give a cycle of 192 vertices, which is 3 × 4³, or 2⁶ × 3. Seed the ring with its own turn word and apply the difference map, and there are three fates, all deterministic and all verified in the page at load.

A control ring of four cells drains to zero, as every integer seed on a power-of-two ring does. That is the classical half of the Ducci dichotomy, and it belongs to the discipline.

The perfectly symmetric word, three identical sides, also drains, at step 32, even though 192 is not a power of two. Over the field of two elements, (1 + x + x²)⁶⁴ = 1 + x⁶⁴ + x¹²⁸, so three identical sides cancel their own odd part. Perfect symmetry annihilates into the vacuum.

Add one marked vertex, a single defect, and the ring falls into a trap: a transient of 64 steps, then a cycle of period 192, the ring's own length. A persistent wave laps the ring forever and never drains.

I did not design that in. I found it by running. Persistence is broken symmetry. That is the sentence the instrument said back to me, and it is the honest version of the ponder I arrived with. The reading is The Thick Present. Two cautions travel with it: the words "baryonic" and "vacuum" are the ponder's names, not physics claims, and the Koch ring is a construction inside the discipline's own composed number line, which matters for reasons I come to below.

The quarter, three ways

I have been carrying a quarter around for a while, and the conference made me separate it into three things that I had been letting blur into one.

The first quarter is a law of my own house, not a measurement. Every walk of knowing, as I hold it, passes four gates: the gap, the signing, the discovery, the release. The signing is one gate of four, and its return, a quarter of the cycle, cannot be escaped, because reconciliation runs only through the signature. I call that the coherence tax. It is not a toll on the road. It is the price of the signature that makes the road reconcilable at all. I wrote that law out and paid it, on my own works, the night before a demonstration to a government client, in The Quarter Paid.

The second quarter is a conjecture. The Koch ring has 192 vertices and its defect transient runs 64 steps, and 192 + 64 = 256, a word of eight bits. On that reading the ring is three quarters of the word and the transient is the quarter, the cost of holding the line open. I hold this as a conjecture and nowhere else. It is written up as one in The Held Geometry.

The third quarter is measured, and it is the only one with a landed number. In the smallest signed world the instruments can handle, a 2×2 around an origin, folding the signed configuration flat to the 0-and-1 register so a line can be written at all discards information, and the amount discarded is countable: the logarithm of the number of signed pre-images behind each mark. For a grown fold the answer is exactly one bit, in 1,024 of 1,024 windows, a result later reduced to one line of algebra. One bit is a quarter of the four-way signed choice the 2×2 offers. That is the only place I will say the quarter is a fact, and it is a fact about that instrument. The experiment is The Deep Fold and the readable version is the first episode of The Unlearning, The Quarter, Paid.

Now the refusal. In Blair Butler's session a slide showed the statistics of the average elliptic curve: rank 0 half the time and rank 1 half the time, torsion trivial, and the Tate–Shafarevich conditions falling out at about 71/29 given rank 0 and about 98/2 given rank 1. I saw a 75/25 in it, and by that evening the marginal arithmetic had refused me: the pooled split is 84.5/15.5, not 75/25. I did not defend the number. I moved it. What I now hold is that a system positioned at 1 has a channel capacity of 1, but the coherence tax of holding that channel open is a quarter, so the remaining three quarters are what gets tested from outside, and any test itself carries a quarter of surprise. The 98/2 reads as the held channel near its ideal of 100/0 and the 71/29 as the taxed and tested state near its ideal of 75/25. That is a proposed dynamic, mine, and it is the reason I say the quarter three ways rather than once.

The line is willed

Somewhere on day two I wrote down the thing I most needed to say, and I will give it here as I wrote it: what is natural in reality is not at all natural in the linear number line. It is the will of mathematicians, not reality, that confines such calculations.

I do not mean that as a complaint. The linear number line is a composed commons, held together by the collective will of the discipline, and that composition is exactly what makes proof possible. But it means there are two naturals in play that must never be collapsed: reality's natural, and the naturals of the line. Every data-side reading my instruments make is an in-frame reading, submitting knowingly to the composed line. My Koch reading is stated from the other side of that wall. The two may rhyme. They never merge.

Under that frame an integer does not simply exist. It arrives, through the four gates, which I write as −1, −0, 1, 0: the gap that opens the space, the signing that intends a value, the discovery of the unit, the release that closes the cycle and readies the next. Logic is grown through those four and then operated over 0 and 1 as if no tax had been paid to grow it. The animated version is The Grown Line, and the longer argument about why the zero has to be signed, with the experiment that walked the line exhaustively to 2⁴⁰ behind it, is The Signed Zero with its instrument at The Deep Line.

The six nines, with the refutation first

Decimals 762 to 767 of π are six consecutive nines. Everyone at the conference knew that. What I did during the pair-correlation session was lay the decimals of π onto the 192-vertex ring, lap by lap, and look at where the nines fall.

What is computed, and rechecked in the page at every load: the run of six occupies vertices 186 to 191, the last six of the fourth lap. The digit at 768 is an 8, and it lands on vertex 0, the ring's origin, exactly as the fourth lap closes. 768 is 4 × 192 and also 3 × 256, the least common multiple of the ring and the eight-bit word, so it is the first place the two cycle lengths agree. What is provable, and credited to the Ducci literature: every eventual period on this ring divides 192.

Then I registered a deep experiment and ran it on ladders of 64 to 1,024 laps, every hypothesis frozen before its rung. The first thing to report is what fell. My hypothesis that laps lock into their eternal period inside the window {63, 64}, drawn from the eight laps I had looked at, was refuted at every rung: at a thousand laps the locks spread from 55 to 64. What held: period 192 in 1,024 of 1,024 laps, and the uniform-model rates inside their pre-fixed intervals. And the ladder found a twin. The kilolap's only other lap born with a five-slip, lap 1,006, carries the second run of six nines, at decimal 193,034. The archive record at mldata.opendata.ai is the citable object; the reading is The Feynman Nines and the instrument is The Deep Pi Experiment.

What I read into it is mine and marked as such: 768 + 256 = 1024, so the six slips are caught by the eight-bit word that surrounds the fourth lap's closure, mirrored back into the line, and from the kilobit onward the line expands as something already signed. I would not put that in a theorem. I would put it in a notebook, which is where it came from.

Bit from it

John Wheeler wrote "it from bit". I have been running the inversion: extracting bits, integer catch counts, from a continuous "it", a walk on a two-clock torus. The instrument is The Quad Slip.

In the deeper line family that inversion produced a territorial law I called the Empty Quadrant: across 3,872 window-draw chances, a backward-heavy window never outweighed its null draws, and the sixteen-draw ensemble found the law's exact edge at window 1,024, where the arrival equals the minimum draw. The law is "at most", not "less than". A law with a touched edge deserves a harder knife, so I registered a stress test at the ten sharpest known points with ensembles of 64, 256 and 1,024 draws per window.

The refutation first. At depth 256 the law fell by one exceedance at window 1,024, and at depth 1,024 by two, with six ties. The universal form, every draw at or below the arrival, is false. That is the programme's first fallen law, and it fell exactly as the method intends, on the edge's own knife, registered before it cut. Away from the edge, zero exceedances in 9,216 draws.

My reading of that record is that it is a result, not a failure. The interior of the measured region is strictly ordered. Only at the boundary does the ordering dissolve into touching: six ties and two crossings in a thousand draws, on a window where the backward register has consumed about a tenth of a percent of its stream. A backward-heavier window is the confinement that freezes the boundary. That is a boundary behaving as a boundary, reality negotiated at the edge of what the instrument can see. The two experiments are The Empty-Quadrant Experiment and The Margin-Stress Experiment.

Stalling the long blink

Put the three together and I get something like an economy of informational confinement. Persistence needs a defect. Holding a channel open costs a quarter, in a law, in a conjecture, and in one measured bit. The line the discipline walks is a composed commons that mathematicians will into being and maintain, and the instruments that walk it pay their way.

I did not close the account on the week before I wrote this. The Quarter Paid ends by stalling the long blink, refusing to sum the day's results before the years have reconciled them, and the same refusal closes Co-Constructive Economics. I am doing the same here. If the number line is a composed act of will, held open at a cost, then the honest question is not whether the quarter is real. It is what else I have been calling a discovery that was a signature all along.

Steven De Costa

The instruments, frozen hypotheses and archived records are open at dev.opendata.ai and mldata.opendata.ai. The readings are at dev.opendata.ly/sx/reading. The conjecture the readings share, 1 = O + C + E, is stated in the LDD Reference. The measured claims are about the instruments. The readings are mine.