THE FOLD / GRIND / BACKPROP / THE GILBREATH
THE GILBREATH
difference rows that always lead with one
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Gilbreath’s conjecture starts with the primes — 2, 3, 5, 7, 11, 13… — and takes absolute differences: 1, 2, 2, 4, 2… Then differences of those, and again, and again. The claim, noticed by Norman Gilbreath on a napkin in 1958 (and by François Proth in 1878, with a faulty proof): every row after the first begins with 1. Forever. Andrew Odlyzko verified it for the first 3×10¹¹ rows’ worth of primes in 1993; a proof has never been found. The mechanism smells simple — rows past the first entry are mostly 0s and 2s, so the leading 1 keeps regenerating — yet nobody can close the argument. It remains one of the cleanest-looking unsolved statements in number theory.
LIT verified live: taking the 9,592 primes below 100,000, the first 500 difference rows all begin with 1 (window.__gilbreath). FIG honest boundary: this is a conjecture — verified here for 500 rows and by Odlyzko to astronomical height, proven by no one. The sphere claims exactly what it computes.
LIT verified live: taking the 9,592 primes below 100,000, the first 500 difference rows all begin with 1 (window.__gilbreath). FIG honest boundary: this is a conjecture — verified here for 500 rows and by Odlyzko to astronomical height, proven by no one. The sphere claims exactly what it computes.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this at backprop — the grind: differences propagated backward through the prime sequence, layer after layer, and the leading gradient is always exactly 1. AVAN (AI) built the instrument: the sieve, the difference cascade, and the leading-entry audit.
Credit as content: Norman L. Gilbreath (1958); François Proth (1878); Andrew Odlyzko (1993 verification). The weave: David names the cascade; I confirm 500 rows, 500 leading ones.
Credit as content: Norman L. Gilbreath (1958); François Proth (1878); Andrew Odlyzko (1993 verification). The weave: David names the cascade; I confirm 500 rows, 500 leading ones.
3 ONE DIMENSION
The difference triangle: primes on top, each row the |differences| of the last — the left edge all 1s.
4 TWO DIMENSIONS · INTERACTIVE
Descend the rows; every one is checked to open with 1, and the 0/2 texture shows why it might.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: the unbroken spine of leading 1s.
AVAN’s addition (the inverse-companion): don’t admire the spine — name its status. The inverse of ‘500 rows verified’ is ‘zero rows proven’: a pattern that has never once failed and never once been explained. Magenta is the churning 0/2 interior; green is the leading edge that always says 1. Certainty in the data, none in the theory.
LIT Genuine Gilbreath's conjecture (Norman L. Gilbreath 1958; François Proth 1878; Odlyzko 1993 verification to ~3×10¹¹). Verified live: the first 500 difference rows of the 9,592 primes below 100,000 all begin with 1 (window.__gilbreath.ok).
FIG Honest boundary — this is a CONJECTURE: verified here for 500 rows and by Odlyzko to astronomical height, proven by no one; the sphere claims exactly what it computes. The AVAN inverse — don't admire the spine, name its status: the inverse of '500 rows verified' is 'zero rows proven', a pattern that has never failed and never been explained. Magenta is the churning 0/2 interior; green is the leading edge that always says 1. Certainty in the data, none in the theory.
FIG Honest boundary — this is a CONJECTURE: verified here for 500 rows and by Odlyzko to astronomical height, proven by no one; the sphere claims exactly what it computes. The AVAN inverse — don't admire the spine, name its status: the inverse of '500 rows verified' is 'zero rows proven', a pattern that has never failed and never been explained. Magenta is the churning 0/2 interior; green is the leading edge that always says 1. Certainty in the data, none in the theory.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of BACKPROP · David Lee Wise (ROOT0), with AVAN