A loop is a test of a world.

Imagine living on a sphere. You can move along its surface, but never leave it. Lay down a loop of string and pull it tighter. It can shrink to a point without cutting through the sphere.

Every loop on a sphere can do this. That turns a familiar shape into a surprisingly powerful question: can we recognize a space by what happens to its loops?

A hole changes the answer.

On a doughnut, this loop winds around the central hole. Pulling it tighter cannot make it disappear. It stays caught around something the surface does not let you cross.

A space is simply connected when it is connected and every loop can contract to a point. The sphere passes this test; the doughnut fails it.

Now add one dimension.

Poincaré asked about a whole three-dimensional space, not a two-dimensional surface. Locally, such a space has three directions to move in, like the room around you.

A three-sphere, written S³, is one example. One way to build it is to take two solid three-dimensional balls and identify their entire boundary spheres. The joined space has no boundary.

The claim: every closed, simply connected three-dimensional manifold is topologically S³. “Closed” means compact and without boundary; “topologically” allows continuous reshaping, not cutting or gluing.

Change the geometry to reveal the topology.

The loop condition tells us which spaces the conjecture concerns. It does not tell us how to recognize an arbitrary one as a three-sphere. We need a way to reveal its structure.

The distances and curvature of a space are its geometry. Its underlying pattern of connections is its topology. The key opportunity is that we can change the geometry without changing the topology.

Richard Hamilton introduced Ricci flow: a rule that evolves the metric—the way distances are measured—according to curvature. The hope is to make a complicated geometry easier to recognize.

During smooth flow the topology stays fixed. The equation changes the geometry intrinsically; it does not require the space to sit inside a larger room.

But the flow can break.

The strategy would be much simpler if Ricci flow always made a space easy to recognize. But curvature can concentrate instead. Watch the waist of the shape: a neck can narrow toward a pinch, where the smooth evolution breaks down.

At that point, we cannot simply keep running the same smooth flow. Nor can we cut the neck away and forget it: cutting can change the very topology we are trying to identify.

To rescue the strategy, we need to understand the singular region well enough to continue—and keep track of what the intervention changes.

The narrowing neck is one useful model of what can go wrong—not the only possible singularity.

The singularities have structure.

The way forward is to show that these apparently troublesome regions have enough structure to be handled. Perelman developed quantities and estimates that constrain how the geometry can degenerate. His non-collapsing results prevent certain uncontrolled losses of volume at the relevant curvature scale.

With that control, mathematicians can zoom into high-curvature regions and analyze standard local models. Among them are necks with a two-sphere cross-section.

The breakthrough was not just spotting a thin neck. It was proving enough control to recognize and handle these regions reliably.

Cut carefully. Keep the record.

Before an appropriate neck becomes singular, surgery removes the neck region and caps the resulting boundary spheres. The flow then continues on the remaining pieces.

Cutting changes topology, so we must keep track of exactly what each operation does. Cuts along two-spheres have a well-understood relationship to connected sums: spaces joined through small openings.

Perelman’s estimates make the surgery process controlled. That is what lets the geometric evolution remain useful as evidence about the original space.

For these spaces, the process ends.

For a closed, simply connected three-manifold, Perelman proved that Ricci flow with surgery becomes extinct in finite time: eventually no components remain.

This is a separate theorem, not something we infer because a drawing shrinks. Its proof controls a geometric measure of families of surfaces and shows that the flow cannot persist forever in this case.

Extinction gives the reconstruction an endpoint. There is no unknown surviving piece left to classify.

Work backward—and use the loop condition.

The surgery record and extinction theorem restrict the original space to connected sums of spherical space-forms (three-spheres with certain points identified) and S² × S¹ pieces (a two-sphere carried around a circle), in the orientable case.

Now the original assumption does its work. S² × S¹ has loops that cannot contract. So do nontrivial spherical quotients. Their fundamental groups—the algebraic records of their loops—rule them out for a simply connected original space.

Only S³ factors remain. Taking connected sums of three-spheres produces another three-sphere.

It was a three-sphere all along.

The argument does not say “this animation looks round.” It says: evolve the geometry, control the singularities, track the surgeries, prove extinction, and reconstruct the original topology.

That is how the absence of trapped loops becomes a classification of an entire three-dimensional space.

Perelman’s 2002–03 work completed the Ricci-flow approach and resolved Poincaré’s century-old conjecture. This is the structure of the proof; the estimates that make each step rigorous are its deep mathematical substance.