Start with a 2×2×2 block of cubes and knock out one corner. The seven cubes left behind make a shape called the chair, and copies of it fill space with no trouble at all. In fact they fill it too easily, since they will happily do so in a repeating pattern. Now take each of the chair's 24 flat square faces and cover it with eight tiny bumps and dents, 192 of them altogether, laid out so that two faces will only sit against each other when the bumps on one drop into the dents on the other. That is Chair44.
The claim made for it is a big one. Copies of this single solid still fill space, but the bumps and dents are supposed to rule out every arrangement that repeats. Squares cover a kitchen floor, and so do hexagons, and both will do it in a pattern that repeats; a tile whose copies can cover everything but can never be made to repeat, however you arrange them, is called aperiodic, and a single such tile is an aperiodic monotile. The first one in the plane was found in 2023. Chair44 is put forward as an answer to the same question in three dimensions.
Every solid in the view is the same chair, in one of the 24 orientations a cube can be turned into, and never its mirror image. That is not because mirror copies are forbidden; the paper allows them as tiles. It is because part of the claim is that the two handednesses never mix: every tiling is supposed to be homochiral, all of one hand, and the patches grown here are built from rotations alone. Colour by Rotation to see which pose each chair is in, or by Supertile to see the eight groups the patch was grown in. The chairs are drawn a little shrunk so that you can tell them apart. Push the gap up to 1 and they close up into a solid block with no space left between them.
There are three ways of drawing the matching rule. True size is the honest one, and it looks like nothing at all, because the pyramids are a ten-thousandth of a cell tall, far too small to see. Enlarged pyramids follows the convention of the paper's own figures, with the bases five times over and the heights exaggerated, so that the bumps and dents actually show. Arrows is Chaim Goodman-Strauss's redrawing of the same rule as a single flat arrow on each face, in three colours: blue meets blue, and green meets red. The colours deserve a sentence. Draw an arrow on a sheet of paper and hold it up to the light from behind: it looks exactly the same, because an arrow has a mirror line down its middle. A whole arrow therefore looks the same from either side of its face and can match a copy of itself. A half arrow, one side of that line, is reversed when seen from behind, so green and red are really one marking seen from its two sides, which is why each may only ever meet the other.
The patches are not built by stacking chairs on a lattice. The chair is a rep-tile, a shape that can be assembled from smaller copies of itself: eight chairs, each turned the right way, fit together into a chair twice the size. So a patch can be grown by substitution. Start with one chair. Replace it with its eight children, replace each of those with eight more, and carry on. The buttons give you 1, 8, 64, 512, 4,096, 32,768 and 262,144 chairs, which fill a chair 1, 2, 4, 8, 16, 32 and 64 cells across. Because every chair is the same solid, even the deep patches cost almost nothing to hold in memory, one matrix and a colour apiece. What gives out first is the number of triangles your graphics card can redraw while you turn the thing round. Past that point the page keeps the depth you asked for but draws the rule more simply, dropping from pyramids to arrows to the bare chair, and tells you so underneath.
Now wind the Built slider, slowly. It takes the patch apart and reassembles it one chair at a time, in the order the substitution produced them, and what you see is not a wall going up sideways. It is a chair filling in, then a bigger chair assembling itself out of chairs like it, then a bigger one again. Colour by supertile and you can watch each block of eight arrive as a chair of the same shape as the one it is sitting inside. That nesting is the heart of the argument for aperiodicity, and it is a good deal easier to watch than to read about.
It is also where the aperiodicity is supposed to come from, by the same argument that works for the hat in the plane. Suppose every tiling has to be built this way. Then any patch you care to point at sits inside a larger chair, which sits inside a larger one still, and so on without end. A repeating pattern is one with a shift that carries it exactly onto itself, and a hierarchy like this leaves no room for such a shift: it would have to respect every level of the nesting at once, and no shift other than zero can manage that.
The page also checks the patch against the paper's atlas of contacts. Two chairs sharing a face can sit in any of 2,388 relative positions if all you ask is that their flat bodies fit; the features cut that down to 44. Every face contact in the patch in front of you is one of those 44, and the line under the view says so. That is a result the page computes each time, not something it takes on trust.
The heights are not there for decoration. Each feature has one of twelve depths, positive for a bump and negative for a dent, and two of them will only mate if a bump drops into a dent of exactly the same depth. Where do the twelve come from? Work out which marks are forced to meet which inside the eight-child dissection, close that set up under further refinement, and insist on a bump wherever a dent has to be. That gives 372 equations, and they split the 192 marks into exactly twelve balanced groups; numbering the groups 1 to 12 gives you the table. Each depth is carried by eight bumps and eight dents, so what one adds the other takes away, and the solid still has volume exactly 7, which this page checks on the tile it draws.
Not fully, and it is worth being clear about that. In the plane the question was open for decades, until David Smith, Joseph Myers, Craig Kaplan and Chaim Goodman-Strauss found the “hat” in 2023. In three dimensions the Schmitt–Conway–Danzer biprism has been known for a long time to tile space only non-periodically, but some of its tilings have a screw symmetry of infinite order, a twist combined with a push along the axis that carries the tiling onto itself, which stops it being aperiodic in the strong sense. Chair44 is presented as a solid that closes that gap.
The claim itself is a preprint, not a refereed result. The machine proof that comes with it is kernel-checked only modulo a named hook for each machine-decided step, and the written geometric arguments stand as exposition rather than proof. So the aperiodicity should be read as claimed, not established. Nothing on this page hangs on it. The solid, the eight-child dissection, the patches, the volume and the 44-contact check are all finite computations, and they are checked here, and against the project's own generator, vertex by vertex.
Computed in your browser by a port of the project's Chair44 engine, checked against it vertex by vertex on every run of the site's tests.