Calculus is the mathematics of change and of the infinite. It took humanity roughly two thousand years to build, cost its inventors a bitter feud, and was attacked as logically bankrupt for over a century before being made airtight. This is the story of why it exists — and why this course rebuilds it, from proofs up. It closes with a tour of where calculus runs in the world today.
How fast is something changing at a single instant — the speed of a falling stone exactly now, not its average over a second? This is the derivative.
What is the area, volume, or total accumulated by a quantity whose boundary or rate keeps curving and changing? This is the integral.
Calculus is the discovery that these two questions — slope and area, the instantaneous and the accumulated — are secretly the same question, inverse to each other. That link is the Fundamental Theorem (Unit 8).
The ancients could measure straight edges and static shapes. Motion and curves defeated them. As early as ~450 BC, Zeno of Elea posed paradoxes — Achilles can never overtake the tortoise, an arrow in flight is motionless at every instant — that made one thing clear: anyone who wanted to talk about the infinitely small, or about adding infinitely many pieces, was walking on thin ice. The warning would echo for two millennia.
Building on Eudoxus, Archimedes perfected the method of exhaustion: trap a curved region between inscribed and circumscribed polygons, then make the gap shrink until the area is pinned exactly. With it he proved the area of a parabolic segment is exactly $\tfrac43$ of an inscribed triangle, trapped $\pi$ between $\tfrac{223}{71}$ and $\tfrac{22}{7}$ (about $3.1408$ and $3.1429$), and found the volume of a sphere.
A circle of diameter $1$ has circumference exactly $\pi$. Archimedes trapped it between the perimeters of an inscribed polygon (a little too short) and a circumscribed one (a little too long), then doubled the sides — $6, 12, 24, 48, 96$ — using only square roots and arithmetic, no sine or tangent tables. Press Double the sides and watch $\pi$ emerge. This is the direct ancestor of the limit you meet in Unit 2.
In a work called The Method — lost for centuries, rediscovered on a scraped-over manuscript in 1906 — Archimedes revealed he first found these results by mentally slicing shapes into infinitely many strips and balancing them on a lever. That is integration. But he distrusted it and never accepted it as proof. Related ideas continued to develop in China, India, the Islamic world, and Europe before a systematic calculus emerged in the seventeenth century.
When Europe turned to measuring the heavens and the motion of bodies, the old geometry buckled and the pieces of calculus appeared all at once:
These and many other contributions put the main pieces on the table. Newton and Leibniz then developed systematic versions of calculus independently.
Isaac Newton, sent home during the plague years of 1665–66, developed his "method of fluxions" alongside his work in physics. In the Principia (1687), he presented largely geometric arguments deriving Kepler's laws from universal gravitation. The underlying ideas were deeply tied to his calculus.
Independently, Gottfried Wilhelm Leibniz built the same calculus from a different direction and gave us the language we still write today: the differential $dx$, the ratio $\dfrac{dy}{dx}$, and the integral sign $\displaystyle\int$ — an elongated S for "sum."
A bitter priority dispute erupted over who invented calculus first. It poisoned relations between British and Continental mathematicians for decades and helped isolate parts of British mathematics from Continental developments. Leibniz's differential and integral notation is the one used in this course.
Calculus worked remarkably well, but its early infinitesimal arguments often lacked a precise foundation. An infinitesimal could be treated as nonzero during a calculation and then discarded as negligible. In 1734 Bishop George Berkeley ridiculed these quantities as "the ghosts of departed quantities." His criticism exposed a genuine gap between the successful computations and their justification.
The rescue came in the 1800s. Bolzano, Cauchy, and Weierstrass replaced the vague infinitesimal with the precise limit — the $\varepsilon$–$\delta$ definition. Riemann made the integral exact, and Dedekind and Cantor built the real numbers themselves. Calculus became analysis: rigorous, proven, certain.
Once it could be trusted, calculus became the native language of every quantitative science. A short arc through modernization:
You can compute derivatives without ever proving a theorem. But the history is a warning written in fire: human intuition about the infinite is unreliable — it was wrong for two hundred years. This course teaches you both to compute and to know exactly why the computation is valid. That is the difference between using a tool and being able to trust it — between a technician and a scientist, engineer, or mathematician whose conclusions can be believed.
Read this prologue once now, and once again on the last day of the course. The second time, you will recognize every name — Archimedes' squeeze, Cauchy's limit, Newton's theorem — because you will have proved their ideas yourself.
Three and a half centuries after Newton, calculus is not a museum piece — it runs quietly underneath most of modern technology. Here is where the ideas from the 14 units actually show up at work today, with the unit each one traces back to. If the proofs ever feel abstract, this is what they are for.
Most modern neural networks — image recognizers and language models alike — are trained with gradient-based optimization: take derivatives of a loss function with respect to many parameters and update them. Backpropagation applies the chain rule (Unit 3) efficiently at enormous scale.
The Black–Scholes equation that prices options is a differential equation; the "Greeks" traders watch are literally derivatives of price (Unit 3). Continuous compounding is $e^{rt}$ (Unit 4), portfolio choice is constrained optimization (Unit 6), and expected payoff is an integral over probability (Unit 7).
Autopilots, thermostats, and robot arms run on feedback control — the "I" and "D" of a PID controller are an integral and a derivative. Signal processing rests on Fourier and Laplace transforms (integrals, Units 7–9), and aerodynamics solves the Navier–Stokes equations by numerical integration.
Drug dosing is modelled by differential equations for concentration over time; the epidemic curves of 2020 came from SIR models — systems of derivatives. CT and MRI scanners rebuild images by inverting integrals. "Flattening the curve" is a statement about a derivative.
The smooth shapes of fonts and car bodies are Bézier curves and splines; physics engines integrate the equations of motion every frame; photorealistic rendering solves the "rendering equation," an integral over all incoming light (Unit 7). Animation is calculus at 60 frames a second.
Rocket trajectories and orbits are differential equations and optimization; GPS works only because it corrects for the time dilation of relativity and computes position from rates of change of signals. The Apollo guidance computer ran on numerical calculus.
A probability is an area under a density curve — an integral (Units 7, 12). Maximum-likelihood and Bayesian methods fit models by setting derivatives to zero (Unit 6). Much of statistics is calculus in disguise, which is exactly why a rigorous calculus background pays off in a data career.
Climate models discretize partial differential equations across the whole atmosphere; power grids, supply chains, and logistics are vast optimization problems (Unit 6). The series and convergence of Units 11 and 13–14 underlie the numerical methods that make all of it computable.
Computers represent real numbers with finite precision, so values such as $\sin(1)$ and $e$ must be approximated. Taylor series (Unit 14) provide one principled route; practical math libraries also use range reduction, carefully chosen polynomial or rational approximations, tables, and other algorithms. Calculus supplies both approximations and ways to bound their errors.
$e=\displaystyle\sum_{n=0}^{\infty}\frac1{n!}=1+1+\tfrac12+\tfrac16+\tfrac1{24} +\cdots$. The first seven terms, through $1/6!$, give $\approx2.71806$. The factorial denominator makes the truncation error shrink rapidly, and Taylor's theorem makes that error quantifiable.
A quick map of where each big idea returns as a tool: