One candidate, several angles.
A 73 second animation of how the article's two checks, at indices 17 and 41, exclude infinitely many indices: every odd multiple of 17 or 41, carried by one composition identity and no search, within the paper's stated Lucas-sequence family. It accompanies the plain-language explainer and the article release, v1.0.0.
Illustration, not verification. Erdős problem 364 asks whether three consecutive positive integers can all be powerful; this partial result does not solve it. Read the exact result and its hypotheses.
Loops by default. Space plays or pauses while the stage has focus; the section buttons jump to each section's start. If your system asks for reduced motion, the piece opens paused on its closing frame.
Eight sections, 73 seconds.
What is on screen in each section, with the narration line. This is the whole piece in prose; the animation adds motion, not claims.
Candidate
One integer X receives two descriptions, which meet in one identity.
One number, X, seen two ways. It's one more than a square, and it's also a step in a repeating sequence.
Question
The identity lifts away and the index line N = 1, 2, 3, … draws in.
So the question becomes: at which steps of that sequence can this happen at all?
Even
Angle 1: the prime 2 sweeps the line and strikes every even index.
First angle: the even steps. Count their factors of two and you always get an odd number, something no square allows. Every even step is out.
Mod 8
Angle 2: the mod-8 phase grid; one phase at 17 and 41 survives the prime 2.
Second angle: remainders after dividing by eight. Odd squares always leave a remainder of one. Half the cases fail at once, but at steps 17 and 41 one case slips through.
Reciprocity
Angle 3: a wheel of local signs over index 17, then 41; the product comes out −1 and both are struck.
Third angle, the hard one. Zoom in on 17. Inside a richer number system, every point of view casts a vote: plus one or minus one. A square would need those votes to multiply to plus one. They give minus one. Seventeen falls, and 41 falls the same way.
Propagation
Composition arcs carry the exclusion from 17 and 41 to the first odd multiples while the camera pulls out.
Now the trick that makes it infinite. Step 17 times any odd number behaves exactly like step 17 itself. So each exclusion jumps forward: 51, 85, 119, and on.
Infinity
The front accelerates: an exponential pull-back while every odd multiple of 17 or 41 is struck and a running count climbs into the thousands.
And it never stops. Every odd multiple of 17, every odd multiple of 41: thousands, then tens of thousands, struck out by the same two checks, none of them tested one by one.
Remains
The count keeps climbing; what a solution must still satisfy, and the closing line.
Whatever survives must still pass every other test; nothing here says a solution exists. One candidate, held fixed, until its descriptions can't agree.
What this animation does and does not show.
- It illustrates Theorems 1.1 to 1.3 of Shifted-square obstructions at indices divisible by 17 or 41 in a family of Lucas sequences (Chey G. A. Loveday, version 1.0.0, 11 September 2026; self-published, not peer-reviewed): two fixed checks, at the fixed indices 17 and 41, and one composition identity that carries them to every odd multiple of 17 or 41.
- It is an illustration, not a verification. The votes on the wheel stand for local Hilbert signs; the animation pictures their product, it does not compute it. The certificates themselves are in the article and its release.
- Nothing here asserts that a solution exists, or that Erdős problem 364 is resolved.
- The surviving indices are not classified by this result. Each of them must still satisfy every other condition the article states; the animation excludes, it does not admit.
- The released Lean development covers integer-Jacobi and recurrence exclusions and propagation, not the human Hilbert-reciprocity, maximality or Pell arguments. Three historical Magma inputs are unrecovered. The v1.0.0 source records these limits; this animation supplies neither missing provenance nor an additional proof check.
Back to the explainerArticle release on GitHubSource at v1.0.0