Matrices Are Transformations
When you first see a matrix, it looks like nothing more than numbers laid out in a box. So everyone tries to memorize it — and that's exactly why it stays confusing forever. A matrix isn't a table. It's an action. More precisely, it's a function that moves every point on the plane at once: it stretches space, rotates it, pushes it sideways. Once this one picture settles in, why matrix multiplication is defined the way it is, and what eigenvalues are even looking at, suddenly get easy. Follow along slowly.
Let's build the feel first. There's a grid on screen. Change the matrix's numbers with the sliders. The whole grid bends and stretches with them. It's not one point moving — the entire plane moves as a single piece. That's what a transformation really is: a matrix is a command that says "reshape this space like this." And here's the important part: the grid doesn't crumple randomly. Straight lines stay straight after the transformation, parallel lines stay parallel, and the spacing stays even. Reshaping space while keeping it this "orderly" is exactly what a matrix transformation does.
So how do you record this whole elaborate transformation with just a few numbers? The secret is surprisingly simple: you only have to track two points. The arrow î that steps one unit right, and the arrow ĵ that steps one unit up. Once you know where those two land after the transformation, you're done. And those "landing spots" are exactly the two columns of the matrix. The first column is where î went, the second is where ĵ went. Look at the matrix and hit the toggle. Those four meaningless-looking numbers were the coordinates "send î and ĵ here" the whole time.
"Okay, but knowing only î and ĵ — what about every other point?" This is the heart of it. You don't need to know the others separately. Any vector on the plane is really just some amount of î plus some amount of ĵ. The point (3, 2), for instance, is 3 of î plus 2 of ĵ. And a transformation preserves that mix. So the transformed (3, 2) is just 3 of wherever-î-went plus 2 of wherever-ĵ-went. Drag a vector around and check. This is linearity. Fix the fate of î and ĵ, and the rest of the infinitely many points follow on their own — that property is why four numbers can steer the entire space.
What if you've done one transformation and you want to do another on top of it? Say, rotate first, then stretch. You multiply the two matrices. So matrix multiplication really means "do one transformation, then the next." Remember being taught to multiply rows and columns in that weird way at school? That strange rule is exactly the calculation for "the result of applying the transformations in sequence." That's why order matters. Rotating then stretching gives a different result from stretching then rotating. Swap the order of the two transformations and watch the picture change. AB being different from BA isn't a rule to memorize — it was obvious all along.
Now let's tour a few classic transformations. Rotation, scaling, shear (the sideways push), reflection. Hit the buttons and you'll see each one corresponds to a single matrix. Every "reshape space" move you know turns out to be a matrix. Finally, press "collapse." Space gets crushed down to a single line. This is a special case, and it's exactly when a value called the determinant becomes 0 — the signal that the transformation squashed space. What this "going flat" actually means is what we dig into in the very next lesson (the determinant).