Zero, One, Glory.
A phone throwing data through open air is impressive. What's actually happening — this whole civilization running on a single yes/no question, asked a trillion times a second — is insane. Part one of two: how the wire actually works.
Start with the thing that should bother you and doesn't, because you've stopped noticing it. You are reading this on a screen that received it from somewhere else, probably without a cable touching it, and that fact alone is already strange enough to sit with: a device in your hand pulled words, in order, correctly, out of thin air. Radio has been doing that since the 1890s, so we've filed it under boring. Fine.
Here's the part that isn't boring. Whatever pulled these words out of the air did it using exactly two symbols. Not twenty-six letters, not a rich vocabulary of tones and pulses — two. On. Off. A voltage or the absence of one. A pulse of light or the absence of one. That's the entire alphabet the modern world is written in, and out of it we've built video calls, bank transfers, satellite photos of your own roof, and every argument that has ever happened in a comment section. One bit of information, repeated at a rate of billions per second, is somehow enough to carry the whole thing. That's the fact this article is actually about. Everything else — wires, waves, glass, satellites — is just the delivery mechanism for that one absurd idea.
The trick is older than computers
The word "bit" was coined in 1947 by the statistician John Tukey, as a lazy contraction of "binary digit," and Claude Shannon picked it up the following year for a paper that quietly invented the field of information theory. Shannon's insight, boiled down past what it deserves, was this: any message, no matter how rich — a photograph, a symphony, a sentence — can be broken into a long enough sequence of yes/no questions that answering all of them reconstructs the message exactly. Is the first pixel darker than medium gray? Yes or no. Is it darker than that? Yes or no. Keep asking. You don't need a machine that understands "darker than medium gray" as a concept. You need a machine that can register yes or no, over and over, unbelievably fast, and never get one wrong.
That's the whole unlock, and it's worth sitting with because it explains why computers are dumb in exactly the way they are. A computer doesn't know what a cat is. It doesn't know what a word is. It knows how to flip a switch and check whether a switch is flipped, at a speed that makes the distinction between "knowing" and "being extremely fast at not knowing" academic. Morse code got there first, in miniature — dots and dashes are a two-symbol alphabet too — which is why the first true test of this idea over real distance was, in 1858, a telegraph cable across the entire floor of the Atlantic Ocean. Queen Victoria's inaugural message to President Buchanan was ninety-nine words. It took sixteen and a half hours to transmit, at a rate of about two words a minute, because the insulation was poor and the cable itself acted like a leaky hose for electrical signal. The line died completely a month later. It would be eight years, a Civil War, and a second attempt before a transatlantic cable actually held.
What a signal actually is
So what is a "bit" doing, physically, while it travels? On copper wire, it's usually a voltage — the line sits at roughly zero volts for a 0 and a few volts for a 1, and a receiver at the other end just checks, thousands or millions of times a second, which state the wire is currently in. On radio, it's a wave that gets switched on and off, or shifted in phase, in a pattern the receiver knows to decode. And on the backbone of the actual internet — the part that makes the whole system possible — it's light.
Fiber-optic cable is astonishingly simple in concept and astonishingly hard to make: a strand of glass thinner than a human hair, so pure that if you built a window out of the same glass, thirty stories thick, you could still see through it. A laser at one end blinks — on, off, on, off — and because the glass is denser than the air around the core, the light physically cannot escape the strand; it just keeps bouncing down the inside of the tube via total internal reflection, arriving at the other end still blinking in the same pattern it left. Light in a vacuum moves at just over 186,000 miles per second. Light inside glass fiber moves at about two-thirds of that, roughly 124,000 miles per second, slowed by the density of the medium — which means a request from New York to London, round trip, spends about 56 milliseconds just being light in a tube under the ocean, before your computer or the server on the other end does a single thing with it. There are more than 870 active submarine cables wrapping the planet, over 900,000 miles of them, and they carry something like 99 percent of all intercontinental internet traffic. Satellites get the headlines. Glass on the ocean floor does the work.
The idea that actually built the internet
None of that explains why the internet works the way it does, though — why it doesn't collapse the instant one wire gets cut, why your video call doesn't fail just because a construction crew in Ohio sliced through a random cable. That part isn't about physics at all. It's about an idea from the early 1960s, developed almost simultaneously and independently by an American engineer at RAND named Paul Baran and a Welsh scientist at Britain's National Physical Laboratory named Donald Davies, who also happened to coin the word for it: packet switching.
Before packet switching, the telephone network worked the way you'd naively design it: circuit switching. When you called someone, the phone system physically reserved one continuous path — an actual dedicated circuit — between your phone and theirs, for the entire length of the call. It worked, but it was brittle and wasteful. One damaged switch anywhere along that single reserved path, and the whole call died. And the circuit sat reserved and idle during every pause in conversation, unusable by anyone else.
Baran's proposal, born out of Cold War anxiety about a network that needed to survive a nuclear strike on any single node, threw that model out. Don't reserve a path. Instead, chop every message into small chunks — packets — and stamp each one with the address of where it's going. Hand each packet to the network and let it find its own way there, hopping from router to router, taking whatever path is open at that instant, with no two packets required to take the same route. At the destination, a receiving computer waits for all the pieces and reassembles them in order, even if they arrived out of order, even if they came by wildly different paths. If one link in the network goes down, packets simply route around it, the way water finds a new channel around a fallen branch. No one packet is precious. No one link is load-bearing. That single reframing — stop protecting the path, protect the message — is the idea the entire internet is a 55-year elaboration on.
The theory got tested for real starting in 1969, when a project funded by the U.S. Department of Defense's Advanced Research Projects Agency — ARPANET — connected a computer at UCLA to one at Stanford Research Institute, 350 miles north. The very first message sent over that link was supposed to be the word "LOGIN." The system crashed after the third letter. The actual first message ever transmitted across what would become the internet was, appropriately, "LO."
The actual first message ever sent over the internet was "LO." The system crashed before it could finish the word.
Finding the address, and agreeing on the handshake
Packet switching solves how a message survives the trip. It doesn't solve where it's going. Every device connected to the internet has an address — an IP address, a string like 142.250.72.14 — the way every house has a street address, and routers along the way read that address off each packet the way a mail carrier reads an envelope, passing it toward a router slightly closer to the destination, then closer, then closer, typically arriving in somewhere between ten and twenty hops, in well under a tenth of a second, for a trip that might cross an ocean.
Nobody memorizes IP addresses, though, which is where the Domain Name System comes in — invented by Paul Mockapetris in 1983, essentially the internet's phone book. When you type a web address, before anything else happens, your computer quietly asks a DNS server "what's the actual numeric address for this name," gets an answer back, and only then sends the real request. You never see this happen. It happens dozens of times during an ordinary five minutes of scrolling.
And once a packet knows where it's going, the two computers on either end still need to agree on manners before any real data moves — which is the job of TCP, the Transmission Control Protocol, designed by Vint Cerf and Bob Kahn starting in 1974. Every time your device talks to a server, it opens with a small, formal ritual called the three-way handshake: your computer sends a packet that essentially says SYN — I'd like to start a connection. The server replies SYN-ACK — acknowledged, I'd like to also start one with you. Your computer replies ACK — acknowledged, go. Three packets, no actual content in any of them, just two machines confirming they're both listening before either says anything real. On January 1, 1983 — a date network engineers still call "flag day" — every single computer on ARPANET was forced to switch over to this TCP/IP handshake system simultaneously, or be cut off from the network entirely. That flag day is, as reasonably as any single date can claim it, the internet's actual birthday.
Where that leaves you
So: this article, the one you're reading right now, was never a "thing" that moved. It was shattered into dozens of numbered packets the instant it left a server, each one stamped with your device's address, each one handed off to whichever router had room at that microsecond, several of them very possibly taking different physical routes through different undersea cables to get to you, arriving out of order, and reassembled, silently and correctly, before your eyes finished the first sentence. It happened at a meaningful fraction of the speed of light, inside glass, on the floor of an ocean, because seventy years ago a statistician needed a shorter word than "binary digit" and a defense researcher wanted a network that a war couldn't kill.
That is, genuinely, insane. You were right to notice it. Most people don't.
Part two, "Please, and Thank You," picks up exactly where this leaves off — because everything above only gets you a pile of correctly-arranged bits sitting on a server somewhere, and a pile of bits is not a webpage, is not a padlock icon telling you your credit card number is safe, is not "the cloud." Next: what actually happens in the quarter-second between typing an address and a page rendering, why "the cloud" is just a polite word for someone else's computer, and how an entire industry of encryption exists to solve one very old problem — that packet switching, for all its genius, was never designed with privacy in mind, and anyone sitting on the right router could read your mail if nobody had fixed that.
Sources and further reading
Historical and technical detail draws on Claude Shannon, A Mathematical Theory of Communication (1948); Paul Baran's RAND reports on distributed communications networks (1960s); Donald Davies' work at the UK National Physical Laboratory; contemporaneous accounts of the 1858 and 1866 transatlantic telegraph cables; the UCLA/Stanford Research Institute ARPANET logs of October 1969; Vint Cerf and Robert Kahn, "A Protocol for Packet Network Intercommunication" (1974); and TeleGeography's public data on submarine cable infrastructure. Figures for fiber refractive index and light propagation speed are standard telecom-industry constants.
Loading.