What is a Computer?

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim." — Edsger W. Dijkstra

You are holding one — in your pocket, on your desk, in the phone that woke you this morning. You use computers fluently and yet, pressed to say what one is, most people reach for a picture of a screen and a keyboard. That is like defining a book by its cover. Underneath the glass sits one of the most consequential ideas humans have ever had, and it is far simpler — and far stranger — than the hardware suggests. This chapter is about that idea, because every other chapter in this Volume is a detail of it.

Core Concepts

A computer is a machine that follows instructions to transform input into output — and, crucially, one whose instructions can be changed. Strip away the screen, the brand, the operating system, and every computer that has ever existed reduces to a four-beat loop: it takes input, processes it by following a list of instructions, produces output, and along the way it can store state to come back to. Input, process, output, storage. A calculator does it. A supercomputer simulating the climate does it. The difference is scale, not kind.

But that definition still describes a washing machine. The word that separates a computer from every other machine is programmable. A washing machine follows instructions too — but they are baked into its wiring. You cannot convince a washing machine to play chess. A computer's instructions are not welded in; they are data you can replace. Change the instructions and the same physical machine becomes a different tool entirely. That single shift — from a machine built to do one thing to a machine that can be told what to do — is the whole revolution.

The stored-program idea

Here is the move that made modern computing possible, and it is subtler than it looks. Early machines drew a hard line: the machine was the hardware, and the instructions came from outside — patch cables you rewired by hand, punched cards you fed in. Reprogramming ENIAC to solve a different problem could take a team of people the better part of two days, physically re-plugging the thing.

The breakthrough was to notice that instructions are just numbers too. If a program is a sequence of numbers, and a computer is a machine that reads and writes numbers, then the program can live in the same memory as the data it works on. This is the stored-program concept, and it is why you can install an app in seconds instead of rewiring a room. The computer fetches an instruction from memory, does what it says, fetches the next, and repeats — millions or billions of times a second. Software stopped being a physical act and became information.

That arrangement — one memory holding both instructions and data, a processing unit that fetches and executes them one after another — is the von Neumann architecture, named for the 1945 report that described it. Nearly every computer you will ever touch is, at its heart, a von Neumann machine. Its parts map cleanly onto our four beats:

What an "instruction" actually is

We keep saying "instructions," so let us be exact, because interviewers will be. An instruction is not a line of Python. It is something far smaller and far dumber: a single, primitive command the CPU understands directly, encoded as a number. Add these two numbers. Copy this value from memory into a register. If this value is zero, jump to that instruction. Compare these two. That is roughly the whole vocabulary — arithmetic, moving data around, comparing, and jumping.

A CPU has no idea what a website, a photo, or your name is. It knows how to add, move, compare, and jump — and it does so astonishingly fast. Everything else — this sentence rendering on your screen, a video call, a neural network — is built out of billions of those tiny operations stacked into towers of abstraction. When you write total = price + tax in a high-level language, a chain of translators (compilers, interpreters) grinds it down into a few of those primitive numeric instructions the silicon can actually execute. The gap between "add two numbers" and "run a video call" is not magic. It is layers — and this whole Volume is a tour up that ladder.

Universality — the idea underneath everything

Now the deepest point, and the one that separates people who use computers from people who understand them. In 1936 — before a single electronic computer existed — a 24-year-old named Alan Turing described an imaginary machine so simple you could explain it on a napkin: a tape of cells, a head that reads and writes symbols, and a tiny table of rules. He then proved something staggering: a machine of this kind, given the right rules, could compute anything that any other such machine could compute. He called it a universal machine.

The consequence is the hidden foundation of the entire industry: hardware does not determine what a computer can do — software does. A computer is not a tool for one job that happens to be flexible. It is a universal machine that becomes whatever program you load into it. Your laptop is a word processor, a guitar tuner, a chess grandmaster, and a film studio — not because it has parts for each, but because it is universal and you fed it different instructions. This is why the software industry exists at all. We do not manufacture new machines for new problems; we write new instructions for the universal one we already have.

Common Mistakes

• Confusing the computer with its screen. The monitor, keyboard, and case are peripherals. The computer is the part that fetches and executes instructions. A server with no screen is still every inch a computer.
• Thinking hardware determines capability. "My phone can't do X" is almost always a software statement, not a hardware one. Universality means the limit is usually the program, not the silicon.
• Believing a program is the text you wrote. Your source code is a human convenience. What runs is a stream of primitive numeric instructions the CPU understands — your code is translated down to them first.
• Treating instructions and data as different kinds of things. In a stored-program machine they live in the same memory and are both just numbers. That's the whole trick, not a footnote.
• Saying "the CPU is the brain" and stopping there. It's a fast, obedient, unthinking clerk that can only add, move, compare, and jump. Intelligence in software is built from those primitives — it is not inside them.

Interview Corner

Chapter Summary

• A computer is a programmable machine: input → process (by following instructions) → output, with storage for state — the four-beat loop under everything.
• "Programmable" is the dividing line: a computer's instructions are replaceable data, not welded-in wiring.
• The stored-program idea — instructions are just numbers living in the same memory as data — is what makes software a thing you can install instead of a room you rewire.
• The von Neumann architecture (CPU + shared memory + I/O, running fetch-decode-execute) is the design nearly every real machine still follows.
• An instruction is a tiny primitive — add, move, compare, jump — and everything richer is layers of abstraction stacked on those.
• Universality (Turing, 1936): one machine can run any program, so what a computer can do is decided by software, not hardware. That is why our whole industry writes instructions instead of building new machines.

Editorial reflection — the temptation is to treat "what is a computer?" as a warm-up question, beneath a real engineer. Resist it. Almost every hard bug, every performance mystery, every security hole you will ever chase is a place where the machine is behaving exactly as a fetch-execute loop over stored instructions must — and your intuition, built on screens and apps, expected something else. Understand the machine as it actually is, and you spend the rest of your career being surprised far less often than everyone around you.