Hypertext
The fact that you can read this from any part of the world because your computer and mine are connected. That everything runs on zeros and ones. And that it all happens at the speed of light, I find that, honestly, extraordinary.
This is my attempt to understand how it works, and a small tribute to the people who changed how we share information, and with it, how the world works.
The paper that invented computing before computers existed
A 24-year-old Turing imagined a theoretical machine that could read and write symbols on an infinite tape, following a finite set of rules. He used it to prove that no general algorithm exists for deciding all mathematical statements, but in the process, described the universal programmable machine. Every computer, server, and smartphone that exists today is a physical instantiation of that idea.
"LO", the first two letters before the system crashed
On October 29, Charley Kline at UCLA typed "LOGIN" to connect to the Stanford Research Institute. The system crashed after "LO." But the network survived, and with it, the idea that computers could talk to each other across distances using packets instead of dedicated circuits. Four nodes in 1969. Millions of networks today.
His CERN supervisor wrote in the margin, "Vague but exciting"
Working at CERN, Berners-Lee wanted a way to manage the institute's tangled web of research documents. He proposed linking them with hypertext, clickable references that could point to any file, on any machine. He wrote the first web server, the first browser, and the first website himself. He never patented it. The web was a gift.
Gil Elbaz, open web corpus
A non-profit crawler that archives the open web and makes the data freely available to anyone. Modest in its moment, essential in retrospect, GPT, BERT, LLaMA, and almost everything that followed was trained on datasets derived from this corpus.
The paper that quietly ended the previous era of AI
Eight researchers at Google published a paper introducing the Transformer, an architecture built entirely on self-attention, replacing the recurrent networks that had dominated NLP for a decade. The paper title was a joke. The implications were not. GPT, BERT, Claude, LLaMA, every major language model today is a descendant of the architecture described in those 11 pages.