The map that
must lie
Peel an orange and try to press the skin flat. It can't be done — the peel splits and tears, every time. The same stubbornness defeats every world map ever drawn: you cannot flatten a sphere without stretching it or ripping it. That impossibility isn't a mapmaker's failure. It's a theorem — Gauss called it remarkable — and it's the reason Greenland looks the size of Africa.
Flatten the world. Watch it break.
Here is the globe, gridded into little tiles, each carrying a small circle. Drag the slider to flatten it — and choose how it's allowed to fail. Stretch: keep the world in one piece, and the tiles near the poles balloon outward, the circles smearing into fat ellipses. Tear: keep the tiles their true shape, and the world splits into orange-peel petals with gaps yawning open toward the poles.
Those are your only two options. One sheet of paper, no stretching, no tearing — impossible. Every map you've ever seen quietly picked one of these two lies.
At the top the circles are honest — equal little disks on the round Earth. Flatten in stretch mode and the polar circles bloat (that's Mercator's lie: Greenland and Antarctica inflated). Flatten in tear mode and the shapes stay true but the planet rips apart. Spin the globe before you flatten.
Curvature is a thing you can't iron out.
Why is flattening impossible? Because the sphere and the plane have different Gaussian curvature — and curvature is intrinsic. That word is the whole point. Intrinsic means an ant living in the surface, who can only measure distances and angles along it and never see it from outside, can still tell the sphere from the plane. It cannot be fooled by bending.
Here's the ant's experiment. On a flat plane, draw a circle of radius r and its circumference is exactly 2πr. On a sphere, walk out a distance r in every direction and the circle you trace is shorter than 2πr — the surface curves the radii toward each other, so the rim doesn't reach as far around. Measure that deficit and you've measured the curvature, from the inside:
Bending paper into a cylinder or a cone never changes any distance measured along the paper — so a cylinder is still secretly flat (K=0), and you can unroll it. But a sphere has K≠0 everywhere, and flattening it would force some along-the-surface distance to change. That's the stretch you saw, or the tear. This is the Theorema Egregium: Gaussian curvature is determined by intrinsic measurements alone, so no flattening — which preserves them — can erase it.
Bite a pizza slice and the crust end stops flopping — pinching a fold (curving one way) forces the slice to stay flat the other way, because the curvature product K must stay zero for a flat dough. The Theorema Egregium is why you can carry pizza without it drooping. Gauss, geometry, and lunch agree.
Every projection sacrifices something.
Since you can't keep everything, mapmakers choose what to save and let the rest distort. Mercator keeps angles true (great for steering a ship on a constant bearing) but pays in area — it inflates anything far from the equator, which is why Greenland (≈ 2 million km²) looks bigger than Africa (≈ 30 million km², fifteen times larger). Equal-area projections keep sizes honest but shear shapes. Equidistant keeps distances true only from one chosen point.
The bookkeeping of this distortion has a beautiful tool: Tissot's indicatrix. Draw a tiny circle on the globe and see what it becomes on the map. If it stays a circle, angles are preserved there; if it's an ellipse, the long axis shows the direction of greatest stretch; if the ellipse has the wrong area, sizes are off. The bloated circles in the flatten demo above are Tissot indicatrices.
Hear the distortion scream toward the pole.
Distortion isn't an abstraction — it's a number at every point, and numbers can be pitches. Here's a flat equirectangular map with a movable probe. The probe's Tissot ellipse shows the local stretch, and the tone you hear rides the area blow-up sec(lat): dead-on and steady at the equator, rising as you climb, and racing toward a shriek as you near the pole, where the distortion runs to infinity.
Drag the probe up and down. You are listening to the singularity that no flat map can avoid — the pole, a single point on the round Earth, stretched across the entire top edge of the page.
Geometer: the area element scales by sec(lat), diverging at the pole. Cartographer: that's why the Mercator map cuts off near the poles — you literally can't print infinity. Musician: a frequency that runs to a scream is a glissando with no top note. The map's lie, the printer's limit, and the rising tone are the same divergence.
The flat map of the universe is wrong too.
The Theorema Egregium is the hinge the whole subject turns on. Once curvature is intrinsic — measurable from inside, immune to bending — you no longer need an outside. You can do all of geometry from within a curved world, which is exactly what you must do when the curved world is spacetime and there is no "outside" to step into. Einstein's gravity is differential geometry taken at its word: mass curves spacetime intrinsically, and we, the ants, feel it as falling.
It also sets up the next act. We've seen curvature refuse to be flattened locally. Next we add it all up: parallel transport already showed a loop's holonomy equals the curvature it encloses, and geodesics showed straight lines converging. Put them together over a whole closed surface and you get Gauss–Bonnet — where all this local stubbornness sums to a single, unbudgeable integer.
After Tristan Needham, Visual Differential Geometry and Forms (Princeton, 2021). A lesson in the Manifold wing of Reverbs.