Encoding the WorldSaying the most with the least — and making it survive the journey

The first scripts were largely pictographic — a sign for an ox, a sign for a house, a sign for "to give." They were expressive but enormous, taking years to master, which kept reading and writing locked inside a small scribal elite. Encoding the world this way meant carrying an entire warehouse of symbols in your head.

The breakthrough, refined by the Phoenicians around three thousand years ago, was the phonetic alphabet: instead of a symbol per word, a symbol per sound. Suddenly a couple of dozen marks could encode any utterance in the language, and literacy became something ordinary people could acquire. This is encoding's deepest lesson, arriving right at the start of the story: power comes not from a bigger set of symbols but from a smaller one that still captures everything — the first great act of compression.

A base is simply how many symbols you run through before rolling over into a new column. Base-10 is a habit of anatomy, not a law of nature. Leibniz saw that base-2 — binary — was the simplest system possible: every number expressible with nothing but 0s and 1s. For two centuries it stayed a philosophical curiosity, an elegant toy.

Then its hidden virtue surfaced. Anything with two clean physical states — on/off, high/low voltage, charged/uncharged, hole/no-hole in a punched card — can store a binary digit reliably, because the two states are easy to tell apart even when a signal is degraded. That robustness made binary the native tongue of machines, from looms and player pianos to transistors. The entire digital world rests on a discovery from Part 1's spirit: that two symbols, used cleverly enough, can encode any number — and therefore any information at all.

To push a message down a wire or across the air, you need a code: an agreed mapping between symbols and signals. The electric telegraph of the 1830s and 40s made this urgent, and Morse code answered with dots and dashes — shrewdly giving common letters short codes and rare ones long, so that ordinary text flew down the line as fast as possible. It was the alphabet's compression lesson, rediscovered for electricity.

Later schemes took a different tack. The Baudot code — a direct ancestor of the bit-patterns inside your computer — gave every character a fixed-length string of marks, clumsier for humans but far easier for machines to send and sort automatically. Each design struck its own bargain between speed, simplicity and reliability. But all rested on one principle: meaning is not in the signal, it lives in the agreed map between sender and receiver. Lose the map, and the most carefully sent message collapses into noise.

Before Shannon, "information" was a fuzzy notion. He made it a measurable quantity, and his key insight was startlingly simple: information is surprise. A message confirming what you already expected carries little; one that resolves real uncertainty carries a lot. He measured it in bits — each bit the answer to one yes/no question — and defined a source's entropy as the average surprise it produces.

Then came the theorem that stunned engineers: every channel has a maximum capacity, a hard rate beyond which you cannot send error-free — but up to which, with clever enough encoding, you can communicate essentially perfectly even through heavy noise. This drew a sharp line between the possible and the impossible for all communication, forever. The craft that began with Phoenician traders choosing letters had become a rigorous science with laws as firm as thermodynamics. Everything digital lives inside the bounds Shannon set.

Shannon's laws set two opposing jobs. Compression strips out redundancy so a message takes fewer bits: Huffman coding, devised in 1952, hands the most common symbols the shortest codes and proves it's the best you can do — Morse's instinct, now mathematically optimal. Error-correcting codes do the reverse, adding structured redundancy so that if noise flips some bits, the original can still be recovered: Hamming codes, from the same years, detect and repair errors all by themselves.

Every robust channel balances the two — squeeze first, then armour. Even biology plays the game: the genetic code carries built-in redundancy that quietly buffers many mutations. From the alphabet's economy to the genome's self-repair, the whole history of encoding converges on a single discipline — say the most with the least, and make sure it survives the trip. But all of this assumes both sides want the message understood. What happens when they don't — when one party knows something the other must never learn? That is the asymmetry of knowing, and it is Part 3.