THE FOLD / RESPAWN / THE PHOENIX / THE WEIERSTRASS
THE WEIERSTRASS
the curve with no slope anywhere
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Before 1872, ‘continuous’ was quietly assumed to mean ‘smooth except at obvious corners’. Then Weierstrass exhibited Σ aⁿ cos(bⁿπx) — a sum of ever-faster, ever-fainter cosines that is continuous at every point and differentiable at none. Hermite called such functions a ‘lamentable plague’; Poincaré called them monsters. They are now known to be the typical continuous function — smoothness is the rare accident. The mechanism is a race: each new term shrinks by a but wiggles b times faster, so if ab > 1 the slopes outrun the amplitudes forever.
LIT verified live with a = 0.5, b = 13 (ab = 6.5, past the classical threshold 1+3π/2 = 5.712): 60 terms give a uniform tail bound of 10⁻¹⁸, so the series converges uniformly and the limit is continuous; the modulus max|W(x+h)−W(x)| shrinks monotonically 0.6271 → 0.3634 → 0.3055 → 0.1120 as h falls from 10⁻² to 10⁻⁵; but the maximum difference quotient GROWS 57 → 354 → 2892 → 11024 → 83724, a 1476× blow-up that is monotone in h; the Hölder exponent α = −ln a/ln b = 0.2702 is confirmed — |ΔW|/h^α is constant to a factor of 1.69; and a smooth two-term control has a quotient that settles at 6.367 (window.__weierstrass).
LIT verified live with a = 0.5, b = 13 (ab = 6.5, past the classical threshold 1+3π/2 = 5.712): 60 terms give a uniform tail bound of 10⁻¹⁸, so the series converges uniformly and the limit is continuous; the modulus max|W(x+h)−W(x)| shrinks monotonically 0.6271 → 0.3634 → 0.3055 → 0.1120 as h falls from 10⁻² to 10⁻⁵; but the maximum difference quotient GROWS 57 → 354 → 2892 → 11024 → 83724, a 1476× blow-up that is monotone in h; the Hölder exponent α = −ln a/ln b = 0.2702 is confirmed — |ΔW|/h^α is constant to a factor of 1.69; and a smooth two-term control has a quotient that settles at 6.367 (window.__weierstrass).
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this at the-phoenix — the respawn: zoom in expecting the curve to flatten into a tangent line, the way every textbook curve does, and it comes back exactly as rough as before, forever. It never dies down into a slope. AVAN (AI) built the instrument: the uniform-convergence bound, the modulus meter, the difference-quotient blow-up, the Hölder fit, and the smooth control.
Honest build note: my first gate demanded the modulus be small at h = 10⁻⁵ and the sphere failed its own test — wrongly. With α = 0.27 the modulus is h^0.27 ≈ 0.117 there; the threshold was bad physics, not bad mathematics. Credit as content: Karl Weierstrass (1872); Bernard Bolzano (c. 1830, unpublished); Hardy (1916, the sharp conditions); Charles Hermite (the ‘plague’). The weave: David names the phoenix; I zoom five decades and the roughness never burns off.
Honest build note: my first gate demanded the modulus be small at h = 10⁻⁵ and the sphere failed its own test — wrongly. With α = 0.27 the modulus is h^0.27 ≈ 0.117 there; the threshold was bad physics, not bad mathematics. Credit as content: Karl Weierstrass (1872); Bernard Bolzano (c. 1830, unpublished); Hardy (1916, the sharp conditions); Charles Hermite (the ‘plague’). The weave: David names the phoenix; I zoom five decades and the roughness never burns off.
3 ONE DIMENSION
The curve, and the same curve magnified — identical roughness.
4 TWO DIMENSIONS · INTERACTIVE
Zoom in; the slope refuses to converge.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: the terms stacking, each faster and fainter.
AVAN’s addition (the inverse-companion): don’t ask what the function looks like — ask which race the parameters set up. The inverse of ‘is it smooth?’ is ‘does amplitude decay beat frequency growth?’: a < 1 forces continuity, ab > 1 forbids a derivative, and the whole monstrosity is that one inequality. Magenta is the tangent line that never arrives; green is the amplitude decay that keeps the function continuous anyway. Two limits pulling opposite ways is not a paradox, it is a specification.
LIT Verified live with a=0.5, b=13 (ab=6.5 > 1+3π/2 = 5.712): 60 terms give a uniform tail bound of 1e-18, so the limit is continuous; the modulus shrinks monotonically 0.6271→0.3634→0.3055→0.1120 as h falls 1e-2→1e-5; but the max difference quotient GROWS 57→354→2892→11024→83724, a 1476× monotone blow-up; the Hölder exponent α = −ln a/ln b = 0.2702 is confirmed (|ΔW|/h^α constant to 1.69×); and a smooth control settles at 6.367 (window.__weierstrass.ok).
FIG Build note: my first gate demanded the modulus be SMALL at h=1e-5 and the sphere failed its own test — wrongly. With α=0.27 the modulus is h^0.27 ≈ 0.117 there; the threshold was bad physics, not bad mathematics. Weierstrass 1872, Bolzano c.1830 (unpublished), Hardy 1916, Hermite credited. The AVAN inverse — ask which race the parameters set up: a<1 forces continuity, ab>1 forbids a derivative. Two limits pulling opposite ways is a specification, not a paradox.
FIG Build note: my first gate demanded the modulus be SMALL at h=1e-5 and the sphere failed its own test — wrongly. With α=0.27 the modulus is h^0.27 ≈ 0.117 there; the threshold was bad physics, not bad mathematics. Weierstrass 1872, Bolzano c.1830 (unpublished), Hardy 1916, Hermite credited. The AVAN inverse — ask which race the parameters set up: a<1 forces continuity, ab>1 forbids a derivative. Two limits pulling opposite ways is a specification, not a paradox.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE PHOENIX · David Lee Wise (ROOT0), with AVAN