0:00 Right now, this video is reaching you through glass. Strands of glass about as thin as a human hair, wrapped in steel and plastic, lying on the bottom of the ocean. Almost every message that crosses an ocean travels this way. Not through satellites. Through the sea floor. The network running through those cables was designed around one idea: there should be no center. No single machine, no single building, no single mistake that could bring it down. On November 18th, 2025, a single file inside one company grew to twice its normal size. Within minutes, X, ChatGPT and thousands of other websites stopped answering.
0:45 So how did a network built to have no center end up with so many? How did a project paid for by the Pentagon become a place where six billion people live part of their lives? And why was the very first message ever sent across it only two letters long? To answer that, we have to go back to a night in 1957, when the sky suddenly had something new in it. On October 4th, 1957, the Soviet Union launched Sputnik, the first artificial satellite. It was the size of a beach ball, and it beeped. For Americans, that beep sounded like falling behind. Within months, the United States created a new agency whose job was to make sure it would never be surprised like that again:
1:31 the Advanced Research Projects Agency. ARPA. ARPA funded big ideas, and some of the biggest were computers. In the 1960s, a computer filled a room, cost a fortune, and every lab that had one guarded it. In 1966, a researcher named Bob Taylor walked into his office at the Pentagon and looked at three terminals. One was connected to a computer in Santa Monica. One to Berkeley. One to MIT. Three machines, three keyboards, three different ways to log in. And none of them could talk to each other. Taylor thought that was absurd. Why not connect them all? He took the idea to ARPA's director, Charles Herzfeld.
2:17 According to Taylor, Herzfeld liked it right away and moved a million dollars out of the ballistic missile defense budget to pay for it. The internet started as a fix for a cluttered desk. You have probably heard a different story: that the internet was built to survive a nuclear war. It is one of the most repeated facts about it. And according to the people who built it, it is wrong. There is a grain of truth. In the early 1960s, an engineer at the RAND Corporation named Paul Baran was asking a terrifying question. If a nuclear strike destroyed part of the phone network, how would the military keep talking? A phone call needed one continuous line. Cut the line, and the call was dead.
3:05 Baran's answer was radical. Do not send a message as one piece. Chop it into small blocks, let each block find its own way through a web of connections, and put them back together at the other end. If one route is destroyed, the blocks simply take another. In Britain, a scientist named Donald Davies came up with the same idea on his own, and gave it the name we still use today: packets. ARPA borrowed the idea. But ARPA's goal was not surviving a war. Herzfeld later said the network was built because big computers were rare, and the researchers who needed them were far away. Even Baran agreed that the nuclear story was a myth. Still, the design kept its accidental superpower.
3:54 A network made of packets has no single path that has to stay alive. By the fall of 1969, ARPA had paid a small company called BBN to build special machines: refrigerator-sized computers whose only job was to pass packets along. One went to UCLA. Another went to the Stanford Research Institute, about three hundred and fifty miles to the north. On October 29th, at ten thirty at night, a student named Charley Kline sat at a terminal in Los Angeles. On the phone with him was Bill Duvall, at the other end. Kline was about to log in to a distant computer, over the new network, for the very first time. He typed L.
4:40 Did you get the L? Yes. He typed O. Did you get the O? Yes. He typed G. And the system crashed. The first message ever sent across ARPANET, the network that would become the internet, was two letters long. LO. About an hour later, they tried again, and it worked. By the end of 1969, four computers were connected. Then the network grew, one university at a time. And in 1971, a programmer named Ray Tomlinson wanted to send a note to a person on another machine. He needed a symbol to separate the person's name from the computer's name. He looked down at his keyboard and picked the one that meant "at". Email became the network's first killer app.
5:27 In many ways, it still is. But by the early 1970s there was a new problem. ARPANET was no longer the only network. There were packet networks running over radio, and packet networks bouncing off satellites. Each one spoke its own language. Connecting them was like mailing a letter through three post offices that all used different envelopes. In May 1974, two researchers, Vint Cerf and Bob Kahn, published a paper with a modest title: A Protocol for Packet Network Intercommunication. What it described was a universal envelope. Here is how it works. Every piece of data travels in a packet with two addresses on it: where it comes from, and where it is going.
6:15 Any network can carry that packet, the same way any truck can carry a sealed box. When the packets arrive, a set of rules checks that nothing is missing, asks again for anything that was lost, and puts everything back in the right order. Those rules became T C P and I P. They did not belong to any one network. That was the whole point. They were the language that networks used to talk to each other. A network of networks. An internet. On January 1st, 1983, ARPANET switched over. Any machine that had not learned the new language by the deadline would be cut off. About four hundred computers made the jump, and some of the people who lived through it got buttons that said:
7:00 I survived the T C P I P switchover. That same year, another problem got solved. Computers were found by their numbers, and humans are bad at remembering numbers. So engineers built the Domain Name System, a giant phone book that turns names into numbers. In 1985, the first dot com name was registered. The internet now had a universal language and a universal phone book. What it did not have yet was anything most people would want to use. In March 1989, a British physicist at CERN, the particle physics lab near Geneva, wrote a proposal. His name was Tim Berners-Lee, and he was frustrated.
7:46 CERN had thousands of scientists, and their information was scattered across computers that could not read each other's files. He wanted documents that could link to other documents, anywhere in the world. His boss wrote three words on the cover. Vague, but exciting. Berners-Lee built it anyway: a language for pages, an address for every page, and a program to read them. On August 6th, 1991, the first public website went online. It explained what the World Wide Web was, and how to make your own pages. Then came a decision that may matter more than the invention itself. On April 30th, 1993, CERN put the web's software into the public domain.
8:33 No patents. No license fees. Anyone could build on it, for free. The internet was the road network. The web was the first thing everybody wanted to drive. Until then, the backbone of the internet in the United States was run by a government science agency, and it was meant for research, not business. In 1991, those limits were lifted. In 1993, a team at the University of Illinois released a browser called Mosaic. It put pictures right next to text, and anyone could install it. One of its creators, Marc Andreessen, moved to California and co-founded a company to build something better. They called it Netscape. On April 30th, 1995, the government's backbone
9:23 was switched off for good, and private companies took over the internet completely. A few months later, on August 9th, Netscape went public. Its shares were priced at twenty eight dollars. On the first day, they touched seventy five. A company that had never made a profit was suddenly worth billions. Wall Street had found the internet, and what followed was a gold rush. Every company needed a website. Every website needed investors. On March 10th, 2000, the Nasdaq stock index peaked at five thousand and forty eight points. By October 2002, it had lost about seventy seven percent of its value. The dot com bubble was over. But the cables, the servers and the habits it had paid for stayed.
10:12 Out of the wreckage came the companies that would shape the next twenty years. Google, founded in 1998, turned the chaos of the web into an index. Social networks turned it into a place where you talk to the people you know. And in January 2007, Apple introduced the iPhone. The internet stopped being a place you went to. It went everywhere with you. According to the International Telecommunication Union, about six billion people used the internet in 2025. Roughly three out of every four humans. And still, about two point two billion people are offline, almost all of them in poorer countries. Which brings us back to the glass on the ocean floor. TeleGeography, a firm that maps these cables, counts more than six
11:02 hundred cable systems, in service or being built. Together they carry almost all of the data that crosses oceans. The most cited estimate is between ninety five and ninety nine percent. Satellites carry only a sliver. On September 6th, 2025, cables near Jeddah, in the Red Sea, were cut. The internet slowed across Saudi Arabia, the United Arab Emirates, Pakistan and India, and Microsoft said its cloud traffic through the region was disrupted. The cause was never confirmed. In the past, ships dragging their anchors across the sea floor have done exactly this. And repairs can take weeks, because a specialized ship first has to find the break. The internet has grown new centers on land, too.
11:48 Much of today's web runs inside a handful of cloud companies. On October 20th, 2025, a hidden timing bug in Amazon's automated system for managing addresses erased the address of a single database service, in one region, in Northern Virginia. Snapchat, Fortnite and banking apps went down for hours. Four weeks later came Cloudflare. A routine database change caused one file, used to spot bots, to double in size. That file was copied to every machine in the company's network, and the software reading it had a size limit. It crashed. Large parts of the web went down with it. And speaking of bots. In 2024, for the first time in a decade, automated
12:35 traffic made up more of the web than humans did. Fifty one percent, according to a report by the security company Imperva. So, how did a network built to have no center end up with so many? Because the protocols kept their promise. Packets still find another way. What became centralized was not the design. It was us. We moved our data into the same few clouds, our traffic through the same few companies, and our continents onto a few hundred cables. How did a Pentagon project become a place where six billion people live part of their lives? Because at every turning point, someone chose to give it away. Open protocols in the 1970s.
13:21 A free web in 1993. An open market in 1995. And the first message was two letters long because the network crashed. Fifty six years later, it still does. The difference is that now, the whole world notices. The internet was never a single invention. It was a set of agreements: how to cut a message into pieces, how to address an envelope, how to give a machine a name. Those agreements are still holding. Everything else is just us, building on top of them. Thank you for watching. If you learned something, like and subscribe for more histories of the technology that shaped our world. And tell us in the comments: what is the first website you remember visiting?
14:08 Next, we will tell the entire history of artificial intelligence. See you there.
