the mathematics of change
Two scary words — derivative and integral — turn out to be two everyday ideas: how fast is it changing right now, and how much has piled up so far. That's the whole subject. Let's earn it.
Speed is easy if you give me two moments: distance covered, divided by time taken. But “right now” is a single instant — zero time passes. In zero time you cover zero distance. So speed right now is zero divided by zero, which means nothing at all.
And yet your speedometer shows a number. A photograph of a falling apple still has a speed baked into it. How can something be changing at an instant, when an instant has no room for change? That paradox stumped people for two thousand years.
Instead of zero time, use a tiny bit of time. Measure the average speed over that little window. Then make the window smaller. And smaller. The averages don't fly apart — they close in on one number. That number is the speed at the instant.
On a graph, average speed is just the steepness of the line between two points. As the points slide together, that line settles into the line that grazes the curve. Steepness of the hill, right at your feet. Drag the point below — then press the button and watch the window shrink.
the coral line is the average over a window; the amber line is the answer at the instant. shrink the window — coral becomes amber.
The derivative is nothing more than the steepness of the curve at a single point — the speed of change at an instant. Move along the curve and the steepness changes; the derivative is the rule that tells you the steepness everywhere.
The genius wasn't the answer — it was the move: approach the impossible instead of arriving at it. That act of sneaking up on a value is called taking a limit, and it is the engine under everything that follows.
In England, a young Isaac Newton — hiding from the plague — invented calculus to describe falling apples and orbiting planets. He called the rate of change a fluxion and used it to explain why the universe moves the way it does. Then he barely published it for twenty years.
In Germany, Gottfried Leibniz arrived at the same ideas independently — and gave us the beautiful notation we still use, the elongated S for a sum and the dy/dx for a rate. The two camps accused each other of theft for decades. Today we credit them both, and mostly write it Leibniz's way.
Long before either of them, Archimedes was already slicing shapes into slivers to find their areas — calculus was waiting to be noticed for almost two millennia.
If the derivative breaks change into instants, the integral glues instants back into a total. How far did the car go? Add up speed × tiny time, over and over. How much water filled the tank? Add up flow × tiny moment.
On a graph, that total is the area underneath the curve. The trick is the same as before: chop it into thin rectangles you can measure, add them up, then make the slices thinner. Slide the control and watch a jagged guess melt into the exact area.
few fat slices = rough guess. many thin slices = the total closes in on the true area. that limit is the integral.
Here is the result so deep they named it the Fundamental Theorem of Calculus. Imagine tracking the area piling up under a curve as you sweep left to right. Ask: how fast is that running total growing right now?
The answer is exactly the height of the curve at that point. Accumulating and finding the rate of change are perfect opposites — like multiplying and dividing. Do one, then the other, and you're back where you started. That single fact lets us solve integrals by running derivatives in reverse, and it stitches the whole subject into one idea.
A neural network learns by asking “which way is downhill?” for its error. The derivative is that downhill direction. Every training step nudges the model down the slope you just dragged.
Training a deep model means chaining millions of tiny derivatives together — the “chain rule” from calculus, run backwards through the network. It is calculus all the way down.
Rockets, interest rates, drug doses, weather, the spread of a disease — anything that changes over time is written and predicted in the language of calculus.
Hover or tap each one.
{{ termBody }}