Canvas 2Dzero dependenciesno blur filter anywhere
Every point of the cat is an emitter. Arrows leave it, cross the aperture, and are added at every
point of the image plane. What you see below is that sum, squared — nothing is filtered,
nothing is smoothed. The cat softens because arrows cancelled.
the image that survives the hole
point spread function (log scale)
the hole, as sampled
not run yet. These compare the arrow sum against the textbook envelopes. The Airy and sinc formulas appear NOWHERE in the rendering path — only here, as the thing being compared to.
Every parameter is in the URL. The same URL renders the same bytes, including the photon speckle — the noise is seeded, not random.
This is classical wave optics. It is also, exactly, the quantum amplitude calculation for a single photon. Both sentences are true and neither is a metaphor: the sum being performed below — add a unit arrow for every path through the hole, rotating at one turn per wavelength, then square the length of the total — is the same arithmetic in both stories. Feynman's arrows and Huygens' wavelets are the same arrows. The maths does not know which story you prefer, and it does not contain anything spookier than addition.
One part is irreducibly quantum, and it is on the photon slider. Turn it down and the picture arrives as individual dots. Light does not dim smoothly; it arrives in lumps, and at low enough rates you can count them. That is not an analogy or a visual device — it is what a detector actually records. The probability of a lump landing at a point is the squared arrow sum at that point. Amplitudes add; lumps land.
What this is not: the cat is not in superposition. Nothing here is entangled, nothing is measured-into-being. The cat is an ordinary incoherently lit object. Every point of it sends its own bundle of arrows through the hole; arrows within one bundle interfere with each other, and different points of the cat simply add their brightnesses. That is why the result is a convolution.
The obvious reading of a dark fringe is "the light got dimmer there". It did not. There is no light there. The arrows arrived and pointed in opposite directions and the sum is zero. Nothing was absorbed, nothing was blocked. Cancellation removed it.
And that removal is the entire reason you can see. Every point of the cat radiates in every direction. Without cancellation, light from the cat's left ear would arrive at every point of your retina at once, along with light from every other point — a uniform glow, no edges, no image, no cat. The reason light from the ear arrives at one place is that everywhere else, the arrows cancelled.
Press cancellation: OFF. That flag keeps every arrow's length and throws away its direction — it adds magnitudes instead of adding arrows. It is not a physical situation; it is what the world would look like if interference were switched off. You get the glow. That is the control condition for seeing.
And the same mechanism, pushed past a limit, is what eats the cat. Narrow the hole and the cancellation that used to confine each point's light to one place stops being tight enough. One mechanism, helping and hurting, with the width of the hole as the dial between them. That dial is the top slider.
A circle gives rings — the Airy pattern — because the aperture edge is the same distance away in every direction, so the cancellation lines up into concentric shells. Turn the aperture down to ~0.05 mm and look at the specular highlights in the cat's eyes: they grow haloes.
A slit is the surprise. It is narrow in x and wide in y, so it cancels loosely in x and tightly in y: the image blurs along one axis only. Look at the bar target under the cat — the vertical bars dissolve while the horizontal bars, a few pixels away, stay razor sharp. Same light, same distance, same everything, and one direction survives. That is the two-slit experiment, still visible, wearing a cat.
The arrow turns once per wavelength. A red arrow turns more slowly along the path, so it takes a
bigger detour to reach cancellation, so the spread is wider — the first null sits at
1.22 λ/D and λ is on top. Hold the aperture and drag the wavelength
from 400 to 700 nm: the cat gets measurably softer as it reddens, by a factor of 1.75.
The whole claim of the page is that the softening is computed, not applied. The rendering path
contains no Gaussian, no box blur, no filter:, and no Airy or sinc formula. It contains
one loop that adds cos(k·r) and sin(k·r) over sample points
inside the hole, where r is a real path length through a real geometry.
Run the checks. They take the resulting pattern and compare it to the textbook envelopes it
was never given, and they report the deviation. One of them also fits the best possible Gaussian to
the computed circular pattern and reports how badly it fails — because a Gaussian has no rings and
this does.