This open problem taught me what topology is

This video explores the century-old topological problem of whether every closed loop contains an inscribed square, focusing on the simpler and more approachable case of inscribed rectangles. The proof reframes rectangles as pairs of points sharing the same midpoint and distance, then maps every pair of points on the loop to a point in 3D space, forming a "wild surface" whose self-intersections correspond exactly to the rectangles we seek. A circular loop initially appears to produce a simple dome, but its hidden self-intersection reveals how collisions can be disguised rather than visually obvious. To generalize the argument, the video constructs a torus from all ordered pairs of loop points, then explains why this torus must be folded by identifying swapped pairs, leading naturally to the Klein bottle as the true space of unordered pairs. Ultimately, the video emphasizes that the Klein bottle here is not a mere curiosity but a practical tool for solving a real problem, and that wrestling with such ideas sharpens mathematical intuition and reveals what topology genuinely is.

• ▶ 0:00 The central unsolved question: does every closed continuous loop contain four points forming an inscribed square? Originally posed by Otto Toeplitz in 1911.
• ▶ 0:46 The video focuses on a simpler version—inscribed rectangles—whose proof naturally leads to the Klein bottle, presented as a real problem-solving tool rather than a curiosity.
• ▶ 1:44 Although there is no known practical application, the proof sharpens problem-solving instincts and offers a deeper insight into what topology actually is, beyond Möbius strips or "rubber sheet geometry."
• ▶ 3:07 An inscribed rectangle on a closed loop is reframed as finding two distinct pairs of points whose connecting segments share the same midpoint and the same length; four such endpoints form a rectangle.
• ▶ 3:53 Every pair of points on the loop is mapped to a point in 3D space using the pair's midpoint (x, y) and distance d—a continuous mapping, so a rectangle exists whenever two different pairs land on the same 3D point.
• ▶ 5:35 The resulting "wild surface" in 3D has self-intersections that correspond exactly to inscribed rectangles; in the example, a continuous curve of self-intersections represents a continuous family of rectangles.
• ▶ 7:16 Every loop drawn on the plane has a unique surface above it, and this surface is intrinsically tied to the loop.
• ▶ 7:24 The proof must apply to any loop, not just a single example, showing that self-intersection is inevitable.
• ▶ 7:28 Proving this requires understanding the surface’s true identity and general behavior—this is the main path to the proof.
• ▶ 7:39 A circular loop produces a deceptively simple dome-like surface that appears to have no self-intersection at first glance.
• ▶ 7:59 The circle must self-intersect because infinitely many inscribed rectangles share the same midpoint and diagonal length, so infinitely many point pairs map to the single top point of the dome.
• ▶ 8:14 "Self-intersection" really means two different pairs of loop points mapping to the same output; the circle shows this collision can be hidden rather than looking like a sheet passing through itself.
• ▶ 9:00 The surface being examined is not a standard function graph, which is a key conceptual point.
• ▶ 9:12 Unlike a typical 3D function graph, its inputs are pairs of points on a loop, not two independent numerical coordinates.
• ▶ 9:18 Its outputs are triplets of numbers (points in 3D space), so the visualization is the set of all outputs: a surface embedded in space.
• ▶ 9:30 Chekhov's principle is introduced: a planted gun must be fired later, meaning important elements must be introduced early to make future developments feel earned.
• ▶ 9:53 The flagged "planted gun" is the earlier claim that cross sections near the curve resemble the curve itself.
• ▶ 10:04 This resemblance is explained: close point pairs map to small z-coordinates and midpoints near the curve, so as pairs collapse, the cross section approaches the original curve.
• ▶ 10:21 All pairs of the form (x, x) correspond to points on the original curve itself, connecting the abstract surface back to the loop.
• ▶ 10:27 The aim is to prove facts about self-intersections using this "wild surface" built from pairs of points.
• ▶ 10:38 The proof will hinge on whether this second "mystery surface" can be embedded in three dimensions without self-intersection.
• ▶ 10:50 Assign each point on the loop a number between 0 and 1, like snipping the loop and flattening it onto the unit interval.
• ▶ 11:06 This gives a nearly one-to-one correspondence, with the single exception that both 0 and 1 map to the same cut point on the loop.
• ▶ 11:21 To preserve continuity, glue the number 0 to the number 1, turning the interval back into the loop—a setup that will matter for analyzing pairs of points later.
• ▶ 11:35 Introduce a second unit interval along a y-axis to represent the second point on the loop, with x-axis for the first point and y-axis for the second.
• ▶ 11:47 A pair of loop points corresponds to a single point in the unit square, whose coordinates encode which ordered pair of points is selected.
• ▶ 11:58 The encoding is almost a continuous one-to-one map, but breaks down at the square's edges because coordinates 0 and 1 refer to the same point on the original loop.
• ▶ 12:14 The unit square representation identifies boundary points: left/right vertical edges are equivalent, as are bottom/top horizontal edges, with arrows indicating orientation to preserve.
• ▶ 12:43 Geometric construction: gluing the blue vertical edges rolls the square into a tube, then gluing the green circular ends produces a torus.
• ▶ 13:08 Reaching the torus is a natural consequence of respecting the topological equivalences of loop-pair data—not a random detour.
• ▶ 13:14 The torus is presented as a natural representation of all possible pairs of points on the loop.
• ▶ 13:23 Naturalness means the mapping is two-way and unique: every torus point maps to a unique pair of loop points, and every pair maps to a unique torus point.
• ▶ 13:32 The association is continuous: small wiggles on the torus correspond to small wiggles in the pair of loop points, with no sudden jumps in either direction.
• ▶ 13:44 The torus is “not quite the right surface” because it naturally encodes ordered pairs, which is more information than the problem needs.
• ▶ 13:56 Treating (a,b) and (b,a) as distinct creates a trivial solution: they share the same midpoint and distance, but that only yields an “infinitely thin rectangle.”
• ▶ 14:17 To prove the existence of non-trivial rectangles, pairs must be treated as unordered—so (a,b) and (b,a) must be considered identical.
• ▶ 14:29 The unit square represents pairs of points; to encode unordered pairs, every point (x, y) must be identified with (y, x).
• ▶ 14:33 The square’s diagonal is the reflection axis: points on one side are mirror images of the other, so the two halves contain duplicate information that must be merged.
• ▶ 14:53 Folding the square along the diagonal performs this gluing, bringing each (x, y) into contact with its counterpart—turning the square into a new surface for unordered pairs.
• ▶ 15:05 Every pair of points (x, x) on the loop maps to a point on the diagonal fold/crease, which is colored red.
• ▶ 15:25 The diagonal fold makes the edge-gluing orientation seem contradictory, so the solution is to cut along another diagonal and add new arrows.
• ▶ 15:51 After gluing the original arrows together, the remaining arrows have reversed orientation, setting up the next construction step.

• ▶ 15:59 The constructed shape is identified as a Möbius strip, made with a single half twist.

• ▶ 16:06 This is not an arbitrary craft project; the Möbius strip arises naturally as the geometric representation of all unordered pairs of points on a loop.

• ▶ 16:18 The correspondence is exact and continuous: each strip point maps to a unique unordered pair on the loop, and vice versa, with small changes matching on both sides.

• ▶ 16:31 The red edge of the Möbius strip is not arbitrary—it carries a specific geometric meaning tied to the loop's original structure.
• ▶ 16:35 The red edge originates from the diagonal in the unit square, i.e. points of the form (x, x).
• ▶ 16:39 The red edge encodes degenerate pairs—where the two chosen points on the loop are actually the same point, listed twice.
• ▶ 16:47 The earlier loop-pair surface is connected back to the Möbius strip, which naturally corresponds to unordered pairs of points.
• ▶ 17:05 This correspondence means there is a continuous function from the Möbius strip onto the loop-pair surface.
• ▶ 17:13 The animation shows every point mapping between the surface and the Möbius strip in both directions.
• ▶ 17:31 The map is not just any map: the edge of the Möbius strip corresponds to points of the form (x, x), so it must be confined to the xy-plane.
• ▶ 17:53 A key topological claim is made: it is impossible to embed a Möbius strip in 3D with its edge lying in a plane without the strip intersecting itself.
• ▶ 18:08 If true, self-intersection means two distinct points on the strip land on the same point, giving two distinct pairs of points on the original curve with the same midpoint and same distance—forming an inscribed rectangle.
• ▶ 18:36 Dan Asimov gave a counterexample: a Möbius strip embedded in 3D whose boundary is exactly a circle lying in a plane.
• ▶ 19:21 This is fatal to the naive claim because the strip's interior goes both above and below the circle, unlike the earlier construction that stays entirely above the xy plane.
• ▶ 19:38 The refined claim: it is impossible to map a Möbius strip into 3D so its edge lies in a plane and its interior lies strictly above that plane.
• ▶ 19:54 The section addresses proving the earlier claim by visually reflecting the Möbius strip, showing that combining it with its reflection creates a new closed surface.
• ▶ 20:10 This construction is equivalent to gluing the edge of one Möbius strip to the edge of another, prompting the explicit question of what surface results.
• ▶ 21:00 After using the rectangular diagram, cutting, and gluing, the result resembles a torus diagram, but the narrator emphasizes it is distinct from a true torus.
• ▶ 21:16 The construction of a Klein bottle is introduced as a workaround for gluing two Möbius strips while keeping orientation consistent.
• ▶ 21:29 The Klein bottle is a "celebrity shape" because it has no clear interior or exterior—any inside point can be moved outside.
• ▶ 21:57 Klein bottles cannot exist in 3D without self-intersection, and this fact becomes the key to proving the inscribed rectangle result.
• ▶ 22:20 A self-intersection in the constructed surface corresponds to two distinct pairs of points with the same midpoint and distance—exactly the condition for an inscribed rectangle.
• ▶ 22:39 The focus shifts to a harder, unsolved problem: proving every closed loop contains an inscribed square, not just any rectangle.
• ▶ 22:52 The proof extends by tracking a third piece of information for each pair of points—the angle of the connecting segment—so that two segments with the same midpoint and length differing by 90 degrees form a square.
• ▶ 23:10 The key intuition is that this new setup requires embedding Möbius strips into four-dimensional space, a correct and central instinct for attacking the problem.
• ▶ 23:23 Green and Lobb (2020) proved that for smooth curves, not only can you always find an inscribed square, but you can find inscribed rectangles of every possible aspect ratio.
• ▶ 24:11 Smoothness matters because every point has a well-defined tangent line, giving clean limiting behavior for the angle between pairs of points; this is why the proof works and why rough curves like fractals remain unsolved.
• ▶ 25:14 Topological objects like Möbius strips are not studied for their bizarreness—they are tools for solving problems, and a Möbius strip represents a whole family of shapes, not any single surface.

• ▶ 0:00 Introduction: The Inscribed Square Problem (0:00 - 3:07) - Introduces the unsolved inscribed-square question, the easier rectangle version, and why topology is a valuable tool.
• ▶ 3:07 Reformulating Rectangles and Building the Wild Surface (3:07 - 7:16) - Reformulates rectangles in terms of midpoint and side length, maps pairs of loop points into 3D space, and shows self-intersections correspond to inscribed rectangles.
• ▶ 7:16 From Examples to the Möbius Strip (7:16 - 16:47) - Uses concrete loop examples, explores the wild surface’s cross-sections, and develops a pair-space viewpoint that naturally becomes a Möbius strip.
• ▶ 16:47 Möbius Strips, Counterexamples, and Klein Bottles (16:47 - 23:23) - Connects the Möbius strip to the earlier surface, addresses a counterexample, glues two Möbius strips into a Klein bottle, and returns to the unsolved square problem.
• ▶ 23:23 Recent Progress and Final Takeaways (23:23 - 27:06) - Covers Green and Lobb’s 2020 smooth-curve result, explains why smoothness helps, and reflects on topology and continuous associations.