In late 1971, in a nondescript office at Bolt, Beranek and Newman in Cambridge, Massachusetts, a 30-year-old engineer named Ray Tomlinson sat between two DEC PDP-10 minicomputers that were connected only through the ARPANET. He typed a message on one machine, addressed it to himself on the other, and pressed send.
The text, he later admitted, was probably something like QWERTYUIOP — a keyboard mash he couldn’t quite remember. It was the first email ever sent between two computers, and the address contained a small typographic decision that would end up printed on hundreds of millions of business cards.
Tomlinson needed a character to separate the name of a user from the name of the machine that user was sitting at. He looked down at his Model 33 Teletype keyboard and picked the @ sign. It was, he said afterwards, the only sensible option — no one used it for anything in computing, it didn’t appear inside anyone’s name, and it already meant, in commercial English, at. He later described the decision as purely practical — he wanted a delimiter that would be trivial for the software to parse and could never turn up in a login name.
Two computers, side by side, and a network built for something else
To understand why this mattered, it helps to picture what electronic mail looked like before Tomlinson touched it. In 1971, computers were large, expensive and shared. Dozens of users logged in to the same mainframe from terminals scattered around a building, and early message programs — MAILBOX on some systems, SNDMSG on the TENEX machines Tomlinson used — already let those users leave notes for one another. But the notes never left the machine. If you wanted to send a message to someone on a different computer, even one in the same building, you walked over, or you telephoned.
The two machines Tomlinson used for the test were only a short distance apart. He later told NPR that the keyboards sat about ten feet apart, close enough that he could wheel his chair from one to the other, type on the first and then roll over to see what had arrived on the second. The only thing connecting them was the network.
The ARPANET, funded by the U.S. Defense Department’s Advanced Research Projects Agency, had been running since 1969 and by 1971 connected roughly fifteen sites across the United States. Its purpose was resource sharing: letting a researcher in one state run a program on a computer in another. Person-to-person messaging was not in the design document.
Tomlinson had been working on a file-transfer utility called CPYNET, and it occurred to him that the same protocol could, with a little rewriting, drop a file into another user’s mailbox on a remote machine.
He grafted SNDMSG onto CPYNET and wrote a companion program, READMAIL, to retrieve what arrived. He pursued it, he later said, “mostly because it seemed like a neat idea.” There had been no directive to go forth and invent email.
Why the @ symbol was lying around unused
The at-sign has a genuinely long history — it appears in merchant correspondence as far back as the sixteenth century, standing for the Spanish arroba, a unit of weight of about 11.5 kilograms. By the twentieth century it had settled into a narrow accounting role in English: five apples @ 10¢.
When engineers standardised the ASCII character set in 1963, the @ was included at position 64 (decimal), partly out of a sense that a computer keyboard ought to carry the symbols found on an office typewriter. That was more or less its only job.
Programming languages of the early 1970s didn’t use it. The TENEX operating system Tomlinson was working on didn’t reserve it. Crucially, it did not appear in anyone’s login name — which was the practical problem Tomlinson had to solve. He needed a delimiter that could never accidentally show up on either side of it. The @ was, in his description, the character that didn’t appear in anybody’s name. The full stop and comma were already carrying meaning in file paths and lists. The @ was, essentially, empty real estate.
There is a small twist that Tomlinson himself pointed out later in life. The @ was free of meaning on his own TENEX system, but not everywhere: on Multics, a rival time-sharing operating system, the @ served as the line-erase character, so typing one would wipe out the whole input line. Tomlinson said he was only reminded of this after the fact. It caused real friction for Multics users, who couldn’t easily type an @ address without their command vanishing, and it took a few years for the convention to settle. By then the @ was already spreading across the ARPANET.
QWERTYUIOP and the message that wasn’t saved
The first message itself has been the subject of gentle mythmaking. Tomlinson consistently said he did not remember the exact text, only that it was a test — probably the top row of the keyboard, possibly something else entirely. He sent several such messages back and forth between the two PDP-10s that afternoon, and none was preserved. He was not documenting a historic moment; he was debugging.
He described the test messages as “entirely forgettable.” In that same 2009 NPR interview he recalled they were often just strings of characters, or occasionally a few phrases from the Gettysburg Address. When he was finally satisfied the program worked, he announced it — using the program itself — in a message to his group explaining how to send mail across the network.
He also did not tell his manager. When Tomlinson showed the working program to his colleague Jerry Burchfiel, he reportedly said, “Don’t tell anyone! This isn’t what we’re supposed to be working on.” BBN was under contract to build the ARPANET; nobody had asked for messaging.
Within a couple of years, though, email accounted for a startling share of the ARPANET’s total volume. A 1973 study by ARPA found that roughly three-quarters of all traffic on the network was email — which was not what any of its architects had planned for.
How one symbol survived every rewrite of the internet
The remarkable thing about Tomlinson’s choice is how little it changed. The underlying protocols were rewritten many times over the next two decades. The original CPYNET-based scheme gave way to SMTP in 1982, which is still, in updated form, the protocol carrying nearly all email today.
Domain names replaced the flat list of machine names in 1984. The ARPANET itself was dismantled in 1990. Through all of it, the syntax user@host stayed exactly as Tomlinson had drafted it on a Teletype in a room in Cambridge.
Email in turn seeded ideas that shaped everything after it. It was the first widely used application on the network to treat human communication, rather than shared computation, as the point. That reframing paved the way for bulletin boards, Usenet, instant messaging and eventually the social web.
The @ symbol travelled with it, and picked up new duties along the way — as a prefix for usernames on Twitter, as a mention on Instagram, as the shorthand for tagging almost anyone anywhere online. In 2010, the Museum of Modern Art in New York added the @ symbol to its architecture and design collection, crediting the mark to Tomlinson — an acquisition of something the museum could never physically hold.
For readers curious about how other quiet decisions built the machinery underneath modern life, Make Tech Easier has traced a similar hinge moment in Douglas Engelbart’s 1968 demonstration of the mouse, hyperlinks and shared-screen editing, and the darker corollary in the 1988 Cornell worm that forced the internet to invent security. Small choices, long shadows.
Tomlinson died in 2016, at the age of 74. He rarely gave long interviews about his role, and he tended to deflect the credit onto the network itself and the colleagues who built it. If pressed, he would point out that email would have arrived anyway, sooner or later, in some form. What he never quite conceded was that the specific character — the little curl of ink now sitting between billions of usernames and their domains — was his. He picked it because it was the one no one else was using. Then everyone was.
