A looser kind of 'same shape'
All the way up this ladder, 'the same shape' has meant something measured. Two triangles were the same when they were congruent — equal lengths and equal angles, a perfect copy. Then similarity loosened that to 'same shape, any size', keeping angles but freeing length. The Erlangen-program rung loosened it further still: affine geometry kept parallelism but dropped length and angle, and projective geometry kept only straightness and incidence. Each step let go of one more thing we used to insist on. Topology is the end of that road — it lets go of almost everything.
Here is the picture to keep in your pocket. Pretend every shape is made of soft, infinitely stretchy rubber. You are allowed to bend it, stretch it, shrink it, twist it, squash it — any continuous deformation at all. You are forbidden exactly two moves: you may not tear (rip a hole or cut the rubber apart) and you may not glue (press two separate spots together so they fuse). Two shapes count as topologically the same when one can be massaged into the other under those rules. That is why topology is nicknamed rubber-sheet geometry: it studies the features that survive all stretching and bending, and only break under tearing or gluing.
What a topology actually is
The rubber-sheet story is the right intuition, but mathematicians need it pinned down with no rubber anywhere. The honest definition rests on a single primitive idea: which points count as near each other. Notice that 'near' is exactly what stretching preserves and tearing destroys — bend the sheet however you like and neighbours stay neighbours, but rip it and two points that were close are suddenly on opposite sides of a cut. So topology is built to capture nearness alone, with no notion of how near (no distance, no '|AB|').
The clever device that captures nearness without distance is the open set. On the line, an open set is one like the interval from 0 to 1 without its endpoints: around every point in it you can still take a tiny step left or right and stay inside — no point is sitting on a hard edge. A topological space is then nothing but a set of points together with a chosen collection of its subsets, declared to be the open ones, obeying a few housekeeping rules: the whole space and the empty set are open, any union of open sets is open, and the overlap of two open sets is open. That short list is the entire foundation. Distance is gone; only the bare pattern of which neighbourhoods sit inside which remains.
A TOPOLOGY on a set X is a collection T of subsets of X (the 'open sets') such that: (1) X and the empty set are in T (2) any union of sets in T is again in T (3) the intersection of two sets in T is in T That is the whole definition. No distance, no angle, no straightness -- only which subsets are declared 'open' (= room to wiggle).
Continuous maps, and the rule that bans tearing
If open sets encode nearness, then the maps we should care about are the ones that respect nearness — the continuous ones. You already met continuity informally as 'you can draw it without lifting your pen', and on the real line as 'small changes in input cause small changes in output'. Topology distils that into one crisp condition stated purely in open sets: a map f is continuous exactly when the preimage of every open set is open. Unpack it and it says precisely what we wanted — points that end up near each other came from points that started near each other; nothing got torn apart on the way.
Watch how this single rule outlaws tearing while permitting all the stretching you like. Take the interval from 0 to 1 and stretch it to twice its length: every nearby pair stays a nearby pair, just farther apart in number — the map and its inverse are both continuous, no problem. Now instead try to tear the interval at its midpoint into two separate pieces. Points that were arbitrarily close astride the midpoint are now flung into different pieces, no longer near; the rule is violated, and the operation is forbidden. Stretching is continuous both ways; tearing is not. That is exactly the rubber-sheet contract, now written in clean mathematics.
Homeomorphism: the equals-sign of topology
Now we can say precisely what 'topologically the same' means. A homeomorphism between two spaces is a map f that is (i) a perfect one-to-one matching of their points — every point on each side paired with exactly one on the other — and (ii) continuous, with (iii) its inverse also continuous. The third clause is the quiet hero: it forbids gluing just as continuity of f forbids tearing. When such an f exists the two spaces are homeomorphic, and to a topologist they are genuinely the same object wearing different clothes.
Let us watch the definition do real work. A circle and a square outline are homeomorphic: shrink the square's corners round and you have a circle, with a clean continuous matching both ways and nothing torn or fused. A solid disc and a solid triangle are homeomorphic for the same reason. But a circle and a line segment are not homeomorphic — and this is the first place the definition earns its keep. Cut one point out of the middle of a circle and what remains is still one connected piece; cut one interior point out of a segment and it falls into two. No homeomorphism can turn a 'stays connected when you remove a point' space into a 'splits in two' space, because it would have to tear. The number of pieces left after a cut is a feature topology can measure.
Invariants: how you prove two shapes differ
There is an asymmetry worth facing head-on. To prove two spaces are homeomorphic you exhibit one explicit homeomorphism and you are done. But to prove two spaces are not homeomorphic you would, on the face of it, have to check that every conceivable map fails — an impossible infinity. The escape is the central strategy of the whole subject: find a topological invariant, a feature that is automatically equal for any two homeomorphic spaces. If two spaces disagree on even one invariant, no homeomorphism between them can exist, and you have proved them different in a single stroke.
You have already, just now, used three invariants without naming them. Connectedness — whether a space is all one piece — is an invariant: a connected space can never be homeomorphic to a disconnected one. The 'how many pieces after removing a point' count is an invariant; it is what cleanly separated the circle from the segment. Compactness, roughly the property of being closed-off and bounded with no escape to infinity, is a third — it is why an infinite line cannot be homeomorphic to a finite circle. Each invariant is a question you can ask both spaces; a single different answer settles the matter for good.
And this is exactly where the rest of the rung is headed, because the invariants get sharper and more powerful. The next guide builds the Euler characteristic chi = V - E + F, a single integer that, astonishingly, comes out the same no matter how you carve a surface into pieces, and that quietly counts a surface's holes. Guide 3 uses it to sort every closed surface by its genus — its number of handles — into one tidy list. Guide 4 introduces the fundamental group, which catches holes by watching how loops can or cannot be reeled in. The whole rung is one long answer to the question this guide poses: *when are two shapes really the same, and how would you ever prove they are not?*
Honest limits of the rubber-sheet view
Before climbing on, fix two honest caveats so the picture does not mislead you. First, topology is not a replacement for the metric geometry you spent this whole ladder building, any more than projective geometry replaced Euclid. It is a coarser lens for different questions. A topologist genuinely cannot tell a big circle from a small one, or a square from a circle — those distinctions are invisible at this resolution. When you care about distance, angle, area, or curvature, you reach back down the ladder for the sharper tools; topology answers only the questions about holes, pieces, and connectivity that survive all deformation.
Second, the soft rubber-sheet language is a faithful guide to intuition, but it is not the proof. The rigorous content lives entirely in open sets and continuous maps — that is why we built the real definition alongside the cartoon. The cup-and-doughnut deformation makes the right answer believable, yet establishing that two surfaces truly are or are not homeomorphic still requires the invariants of the coming guides, and some genuine theorems of the rung (the full classification of surfaces, the fact that a sphere and a torus differ) need real work to nail down. The rubber sheet tells you what to expect; the open sets and invariants make it certain.