How to read it

It is a zoom, not a scroll. The view never moves sideways — the frame stays over the same point and only ever gets wider, so everything grows or shrinks about the middle of the window. Which means the thing you were just looking at does not leave: it collapses toward the centre and turns into a speck while the next thing opens up around it. That collapse is the exhibit.

Rungs fade in as you reach them. Two pairs on this ladder are very close — the Sun is only 1.8 times the width of the Moon's whole orbit — and drawn at full strength the bigger one is a wall of light behind the rung you are actually on. So the next rung up stays dark until you travel toward it, and everything you have already passed stays at full strength.

Things overlap, and that is the geometry. In a zoom that converges on one point, everything smaller ends up inside everything bigger. Some of those are literally true — the Earth is inside the Moon's orbit, the Sun is inside the planets' orbits — and some are just a size comparison sitting on top of itself. The Moon's entire orbit fits inside the Sun with room to spare, and on this page it does exactly that.

The calipers bracket exactly the dimension the caption is quoting — the width of the lakes, the height of the tree, the diameter of a sphere, the gap for a distance. When an object is wider than the window the bracket runs off both sides and its ends become arrows, because a caliper that quietly stops at the edge of the screen is measuring the screen.

The grid is powers of ten, in real metres. As you zoom out the lines slide together, fade, and are replaced by the next decade up, which springs back to full spacing. It is there for the long empty stretches: between Stephenson 2-18's comet shell and the Milky Way there is nothing at all to look at for four decades, and without the grid that reads as the page having frozen instead of as an enormous amount of empty space.

The background belongs to the scale. There is no one star wallpaper following you from the electron to the edge of everything. Below a millimetre the backdrop is suspended matter, because a virus and a blood cell are in water; from the Great Lakes to the Moon's orbit the landscape falls away and stars come up in three layers at different depths; from the nearest star the space between things stops being empty and becomes gas and the band of our own galaxy; by the Local Group there are no stars left in the sky at all, only galaxies; and past Laniakea those resolve into filaments. The star layers converge on the middle of the frame at different rates, which is parallax — pointed along the axis this exhibit actually travels on, since the camera only ever zooms.

Every rung has an idle loop, and each one is a motion the thing really makes. The atom's cloud shifts and does not orbit. The elephant fans an ear, which is what an elephant standing still is mostly doing. The comets in the Oort Cloud are on real Kepler orbits and crawl at aphelion. The Sun throws an eruption every seventeen seconds, at a new bearing each time. Arrive at a rung and it reacts once, then settles into something slow enough to sit and watch.

Why the order looks wrong

The ladder is sorted strictly by size, with no exceptions, and that produces one result people argue with: the largest star anyone has measured is smaller than the planets' orbits. So Stephenson 2-18 lands between the Sun and Neptune's orbit rather than out past the Oort Cloud where a story about "the biggest star" would want to put it.

It is not a bug and it is the most useful thing on the page. A star is a small object in an enormous amount of empty space, and the empty space wins every time you zoom out. The Oort Cloud and the nearest star are the same trap read the other way round: the cloud's frame is nearly as wide as the frame holding the whole distance to Proxima, which looks like the Sun's comets almost touching another star. They do not. One of those numbers is a width and the other is a one-way trip, and halving the first to get the cloud's reach puts the comets a bit over a third of the way. Two frames the same size are not two things the same shape.

Some rungs are objects and some are distances — the Moon's orbit, the gap to Proxima Centauri. Those say so in the caption, because a distance quietly presented as an object is a lie told by layout.

The rung after Proxima exists because of that same trap. Zoom out from Stephenson 2-18 and the next two frames open up around the spot it just left, so the Oort Cloud arrives looking like the hypergiant's — you have only just been shown the largest star anyone has measured, and the eye hands the shell to it. It is the Sun's. The cure is one frame further out: Stephenson 2-18's Oort Cloud, the shell that star would actually hold, two and a half times wider and a frame of its own. Seeing the Sun's cloud sitting inside it at four-tenths of the width is the end of the confusion, and it costs the page its only unmeasured rung to buy — see below.

Where the edges are

The geometry is the data. Every position, every size, every gap and every caliper on the field comes from the metre value in the table below, divided by the width of the current frame. There is no "about this big" anywhere in the layout.

The pictures are illustrations. The sizes come from the numbers in the table; the drawings inside them are drawn by hand and their colours are chosen. An elephant is grey because elephants are grey, not because anything was sampled. Treat the shapes as labels on correctly-sized boxes.

Nothing is drawn inside a thing that has no measured inside. This rules out three pictures you have seen a hundred times, and all three are wrong the same way — they draw trajectories for things that do not have positions. The electron has no rings, no shells and no axis, because it is a point particle with no known parts. The hydrogen atom has no ball going round it; what shifts there is a probability of being somewhere, which is the only thing an electron in an atom actually has. And the proton has no three coloured dots: it has three quarks, but a quark is confined, has never been seen alone, and cannot be put anywhere in particular. What is drawn instead is the thing that was measured — the proton's charge density, falling off exponentially the way its dipole form factor says it must, with the quoted radius marked as the moment of that curve it really is rather than as the wall it is usually mistaken for.

The rail is not perfectly logarithmic. It is mostly divided by decades — the three empty decades between the neutrino and the proton genuinely cost you three decades of dragging, and you feel them. But two pairs on this ladder are very close together (you and an elephant; the Oort Cloud and the nearest star), and a strictly logarithmic rail gives those a slice about two pixels wide, which is a rung you cannot stop on. So every gap gets a flat extra 0.45 decades. Across the whole rail that is 10 decades of padding against 45 real ones.

The first three rungs are not measurements. A quark, an electron and a neutrino have never shown a size to anybody. They are point-like in every theory in use, and the numbers this page draws them at are upper limits — the width below which an experiment would have seen no structure anyway, which is a statement about the apparatus and not about the particle. All three are drawn with a dashed edge, and the dashes mean we do not know where this stops.

Which means the order of those three is a fact about experiments, not about particles. The quark's bound is the tightest because the machines that set it reach the furthest — HERA and then the LHC, probing for structure down near 10⁻²⁰ m. The neutrino's is the loosest, because the only size-like quantity anybody has ever bounded for a neutrino is its charge radius, and that comes from watching reactor antineutrinos scatter off electrons, which is a far blunter instrument. So the neutrino sits above the electron here. Nobody thinks it is bigger. Read the three of them as one rung with three names and a note about how hard each was to corner.

And "measured" is not one thing either. Some rungs above the proton are measurements in the plainest sense — the proton's charge radius, the Earth's diameter, Hyperion's height, the distance to Proxima. Others are estimates that depend on a model, or on somebody deciding where a thing stops. Nobody has ever detected an object in the Oort Cloud. Stephenson 2-18's size hangs on a cluster membership that is disputed. The Milky Way has no edge, only a place where the stars run out. Laniakea is defined by which way galaxies are moving rather than by a boundary you could cross. The observable universe's width comes out of a cosmology, not off a ruler. Every one of those rungs carries a caveat saying so, and this page would rather you read them than take the drawing's confidence at face value.

And one rung is not measured at all. Stephenson 2-18's Oort Cloud has never been observed, inferred, or looked for. No object has ever been detected around any star but the Sun. That number is the Sun's own cloud multiplied by the cube root of a mass — a Hill radius, the distance at which the Galaxy starts pulling harder on a comet than its star does, goes as M^1/3 — and the mass is itself an estimate for a star whose distance is disputed. It also assumes the Galaxy pulls out there the way it pulls here, and it does not: Stephenson 2 sits far closer to the Galactic centre, where the tide is stronger, which would draw the shell in by a third or more. It is here because the frame before it is misread by almost everybody, and the only way to show that the Oort Cloud is the Sun's is to put the other star's beside it. The drawing says what it is — the rim is dashed, the way the electron's is, and nothing is falling through it, because nothing has ever been seen to. Treat that one rung as arithmetic wearing the costume of an object.

Things drawn at true scale that you therefore cannot see. The nucleus inside the hydrogen atom is a hundredth of a pixel. The planets on their orbits are thinner than the lines of the orbits themselves. The Sun disappears entirely at Stephenson 2-18's frame, and Stephenson 2-18 itself disappears at the Oort Cloud's — a ten-thousandth of that width, about a seventh of a pixel, and it is left undrawn rather than inflated into something you could see. All of those are correct, and all of them are the reason the diagrams you have seen elsewhere are wrong.

Two things on those last frames are inflated, and both are admitted here rather than in the picture: the glint at the middle of the Oort Cloud and the red one at the middle of Stephenson 2-18's are markers standing in for stars far too small to draw — the Sun there is a twenty-millionth of the frame. They are the same size as each other on purpose, so that the only difference between those two frames is the colour of the star the shell belongs to.

The Great Lakes are traced from longitude and latitude and projected the way a map is, so the shape you recognize is a shape the numbers produced. It is a coarse trace — about sixty points for five lakes — not a coastline dataset.

Three figures, everywhere. No number here is quoted more precisely than that, because the fourth figure is noise on most of these and completely meaningless on some of them.

The ladder

Every rung, and the number the drawing was built from.

Sources

The three bounds at the bottom are limits, not measurements, and each comes from a different kind of machine: HERA's electron–proton scattering and later LHC dijet searches for the quark, Penning-trap and scattering limits for the electron, and reactor antineutrino–electron scattering — TEXONO at Kuo-Sheng, and the accelerator experiments before it — for the neutrino's charge radius, which is bounded at order 10⁻³² cm² in ⟨r²⟩ and is quoted here as the round number that is, about 2 × 10⁻¹⁸ m. CODATA 2018 for the proton charge radius and the Bohr radius. IAU 2012 for the astronomical unit; IAU 2015 for the nominal solar radius. Fok et al. (2012) for the radius of Stephenson 2-18, whose cluster membership — and therefore its size — is disputed; Davies et al. (2007) and the same paper for the Stephenson 2 red supergiants' masses, from which the fifteen-solar-mass figure behind that star's hypothetical comet shell is taken. The shell itself has no source, because there is nothing to cite: it is the Sun's Oort Cloud times the cube root of that mass ratio, which is the standard Hill-radius scaling against the Galactic tide, and it is the only number on this page that came out of a calculation instead of an observation. Hipparcos and later parallaxes for Proxima Centauri, and NASA's mission status page for how far Voyager 1 has got along that line. Bland-Hawthorn & Gerhard (2016) for the shape of the Galaxy. Sawala et al. (2025), Nature Astronomy, for the Milky Way–Andromeda merger being roughly a coin toss over the next ten billion years rather than the settled four-and-a-half-billion-year collision the older simulations gave — the uncertainty is in Andromeda's transverse velocity, and later work using newer Gaia proper motions leans toward a merger without closing the question. Tully et al. (2014), Nature, for the Laniakea supercluster and its definition by velocity flow. Planck 2018 for the age of the universe and the comoving distance to the particle horizon. The shell at the observable universe's rim is real data: the WMAP nine-year all-sky map of the cosmic microwave background, NASA / WMAP Science Team, taken from the nine-year basic results paper and unwrapped from its Mollweide ellipse back onto the sphere it was flattened from. The colours in it are the temperature scale those maps conventionally use — ±200 microkelvin against a 2.7 kelvin background — and not anything the eye could see, since 2.7 kelvin is microwave. The tallest-tree measurement is the 2019 remeasurement of Hyperion. Great Lakes volumes from the EPA and Environment and Climate Change Canada.