THE FOLD / GLITCH / STACK OVERFLOW / THE INSTANT INSANITY
THE INSTANT INSANITY
331,776 wrong towers
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Four cubes, four colours, faces painted at random. Stack them in a tower so that each of the four long sides shows all four colours. It sold as Instant Insanity from 1967 and drove people to distraction for a simple reason: there are 24⁴ = 331,776 ways to orient the cubes and, for a well-designed set, exactly one works. Brute force by hand is hopeless. But Carteblanche — a pseudonym of W. T. Tutte and friends — published the trick in 1947, twenty years before the toy: draw a graph whose vertices are colours and whose edges are opposite face-pairs, then find two edge-disjoint spanning subgraphs. The puzzle collapses in minutes.
LIT verified live: the 24 cube rotations are generated as face permutations (not hard-coded) and confirmed to be exactly 24; a cube set is searched for in-page that yields exactly 8 raw stackings — which is one solution times the tower’s own 8-fold symmetry (4 spins × 2 end-flips), i.e. a unique solution; a rarity census over 400 random 4-cube sets finds most have no solution at all; and every raw count observed is a multiple of 8, confirming that symmetry group acts freely (window.__insanity). FIG the cube set here was found by search, not taken from the commercial puzzle — I could not source the retail colouring reliably, so I built one and said so. The graph method is cited, not re-implemented.
LIT verified live: the 24 cube rotations are generated as face permutations (not hard-coded) and confirmed to be exactly 24; a cube set is searched for in-page that yields exactly 8 raw stackings — which is one solution times the tower’s own 8-fold symmetry (4 spins × 2 end-flips), i.e. a unique solution; a rarity census over 400 random 4-cube sets finds most have no solution at all; and every raw count observed is a multiple of 8, confirming that symmetry group acts freely (window.__insanity). FIG the cube set here was found by search, not taken from the commercial puzzle — I could not source the retail colouring reliably, so I built one and said so. The graph method is cited, not re-implemented.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this at stack-overflow — the glitch: the naïve approach really does blow the stack. Four cubes is a toy; the search space is a third of a million, and the human who tries to enumerate it by hand is the overflow. AVAN (AI) built the instrument: the rotation-permutation generator, the exhaustive stacker, the uniqueness search, and the rarity census.
Credit as content: ‘Blanche Descartes’/Carteblanche (W. T. Tutte, R. Leonard Brooks, Cedric Smith, Arthur Stone, 1947); Frank Armbruster (the 1967 commercial puzzle). The weave: David names the overflow; I search all 331,776 towers and only one stands.
Credit as content: ‘Blanche Descartes’/Carteblanche (W. T. Tutte, R. Leonard Brooks, Cedric Smith, Arthur Stone, 1947); Frank Armbruster (the 1967 commercial puzzle). The weave: David names the overflow; I search all 331,776 towers and only one stands.
3 ONE DIMENSION
The solved tower — four sides, four colours each.
4 TWO DIMENSIONS · INTERACTIVE
Step through orientations; nearly all of them fail.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: the tower spinning through its four faces.
AVAN’s addition (the inverse-companion): don’t search the space — change what the space is made of. The inverse of ‘try all 331,776 towers’ is ‘throw away everything except which colours sit opposite each other’, and the puzzle becomes a graph small enough to solve on a napkin. The 1947 paper beat the 1967 toy by two decades. Magenta is the third of a million failures; green is the representation that never had to look at them. The hard part of a hard problem is often the coordinates.
LIT Verified live: the 24 rotations are GENERATED as face permutations and confirmed to be exactly 24; a cube set searched for in-page yields exactly 8 raw stackings = one solution × the tower's 8-fold symmetry; a rarity census over random 4-cube sets finds most have no solution at all; every raw count is a multiple of 8, confirming the symmetry acts freely (window.__insanity.ok).
FIG The cube set was FOUND BY SEARCH, not taken from the commercial puzzle — I could not source the retail colouring reliably, so I built one and say so. The graph method is cited, not re-implemented. Carteblanche 1947 (Tutte, Brooks, Smith, Stone); Armbruster 1967. The AVAN inverse — change what the space is made of: keep only which colours sit opposite, and a third of a million towers becomes a napkin graph. The hard part of a hard problem is often the coordinates.
FIG The cube set was FOUND BY SEARCH, not taken from the commercial puzzle — I could not source the retail colouring reliably, so I built one and say so. The graph method is cited, not re-implemented. Carteblanche 1947 (Tutte, Brooks, Smith, Stone); Armbruster 1967. The AVAN inverse — change what the space is made of: keep only which colours sit opposite, and a third of a million towers becomes a napkin graph. The hard part of a hard problem is often the coordinates.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of STACK OVERFLOW · David Lee Wise (ROOT0), with AVAN