SOURCE MATERIAL no. 003 — first light

Ten times something crossed a wire for the first time: the first telegraph demonstration, the first memo of the internet's design conversation, the first email, the first spam, the first emoticon, the first commercial domain, the first web page, the first camera, the first text message, the first tweet.

Each panel is the verified record itself — and where the record is lost, the loss is the artifact, in the words of the person who lost it.

compiled, edited & published by Steven Delpercio · delpercio.dev · 13 July 2026


WHAT HATH GOD WROUGHT

AMERICAN MORSE — 1844

Four words from the Book of Numbers, chosen by a young woman, tapped from the Supreme Court chamber to a Baltimore railroad depot — the opening line of electric communication.

WHAT HATH GOD WROUGHT

Samuel Morse had lobbied Congress for years; thirty thousand dollars built the experimental line along the Baltimore & Ohio right-of-way. For the public demonstration on 24 May 1844 he let Annie Ellsworth, daughter of the patent commissioner who had brought him news of his funding, choose the text. She chose scripture: Numbers 23:23, a verse that arrives as a question and reads as astonishment.

At the Washington end, Morse's key encoded the words in the dots and dashes of the original American Morse; forty miles away Alfred Vail's register embossed them onto a paper strip, and Vail sent the same words back to prove the circuit ran both ways. The receiving tape survives at the Library of Congress — the medium's first message is also its best-provenanced, kept from the day it was made.

Within a decade wires shadowed every rail line and news moved at current speed; markets, war, and diplomacy followed. But the demonstration's deepest legacy is the genre it founded: the first message as ceremony, a text chosen for the ages before anyone knows what the medium will be for. Every first transmission since is, knowingly or not, a reply to this telegram.


Request for Comments: 1

MEMORANDUM — 1969

Months before the ARPANET carried a single packet, a graduate student typed a memo so unassuming it apologized in its title — and founded the document series the internet is made of.

Network Working Group                 Steve Crocker
Request for Comments: 1                         UCLA
                                        7 April 1969

                    Title:  Host Software

The software for the ARPA Network exists partly in
the IMPs and partly in the respective HOSTs.

In April 1969 the ARPANET existed as contracts and delivery dates: the first packet-switching node would not reach UCLA until September. The graduate students assigned to make host computers talk to it called themselves the Network Working Group, and Steve Crocker drafted their first memo — numbered 1 — describing how hosts and IMPs, the network's minicomputers, would divide the work.

Crocker later recalled composing it at night and choosing the series title so the notes would not sound like decrees from authority nobody had: Request for Comments. The modesty was engineering: an invitation scales better than a standard. The numbered conversation the memo opened never ended — every core protocol of the internet, from mail to the web's transport, was designed in its pages.

RFC 1 matters less for its content — soon superseded — than for its register. It established that the network would be designed in public, by whoever showed up, in documents anyone could answer. The internet's architecture of openness is, at bottom, a correspondence habit, and this is the first letter.


QWERTYUIOP

ARPANET MAIL — 1971

The most consequential transmission of its century has no surviving text. What survives is its author's refusal to invent one.

The test messages were entirely forgettable and I
have, therefore, forgotten them. Most likely the
first message was QWERTYUIOP or something similar.

                                   -- Ray Tomlinson

In late 1971, Ray Tomlinson of Bolt Beranek and Newman joined two programs — SNDMSG, which left messages in mailboxes on one machine, and CPYNET, which copied files between machines — so that a message could land in a mailbox on a different host. He addressed it with a character no username contained, and user@host became the grammar of reaching anyone.

The first messages went from one PDP-10 to another standing beside it, connected through the ARPANET rather than the room. They were tests, sent to himself; nobody present recorded them, because nothing about the day announced itself as history. Asked ever after what the first email said, Tomlinson gave the answer in the panel: forgettable, forgotten, most likely the top row of the keyboard.

The lost text is the negative space of this history. Between Morse's engraved scripture and Dorsey's timestamped shrug sits the medium that carried more human correspondence than any before it — opening with words no archive holds. Tomlinson's honesty, maintained against fifty years of interviewers hoping for poetry, is the artifact: where the record ends, the right answer is unknown, not a better story.


DIGITAL WILL BE GIVING A PRODUCT PRESENTATION

ARPANET MAIL — 1978

A computer salesman with a printed network directory and a deadline invents, in one afternoon, the inbox's oldest complaint.

DIGITAL WILL BE GIVING A PRODUCT PRESENTATION OF
THE NEWEST MEMBERS OF THE DECSYSTEM-20 FAMILY; THE
DECSYSTEM-2020, 2020T, 2060, AND 2060T. ...

WE INVITE YOU TO COME SEE THE 2020 AND HEAR ABOUT
THE DECSYSTEM-20 FAMILY AT THE TWO PRODUCT
PRESENTATIONS WE WILL BE GIVING IN CALIFORNIA THIS
MONTH.

Gary Thuerk of Digital Equipment Corporation wanted West Coast ARPANET users to know DEC's new machines spoke their protocols natively, and saw no reason to send hundreds of individual invitations. On 3 May 1978 his message — typed in the upper case of the sending system, addressed from the printed network directory until the header overflowed — invited the research community to product presentations in Los Angeles and San Mateo.

The network's stewards reacted as if a stranger had pinned an advertisement inside a laboratory. The Defense Communications Agency phoned DEC to extract a promise it would not recur; the discussion thread that followed — preserved along with the message — rehearsed, fully formed, every argument about unsolicited commercial messaging the next half century would repeat. One participant allowed that he wouldn't mind announcements he happened to find interesting.

The word spam lay fifteen years in the future, borrowed eventually from a Monty Python sketch by way of early chat systems. The phenomenon needed no name to work: the presentations were well attended, the sale by Thuerk's later account enormous, and the economics — advertising whose cost falls on the recipient — proved as durable as email itself.


:-)

CMU BBOARD — 1982

Eleven forty-four in the morning, 19 September 1982: a computer scientist proposes three characters to mark a joke, and text acquires a face.

19-Sep-82 11:44    Scott E  Fahlman             :-)
From: Scott E  Fahlman <Fahlman at Cmu-20c>

I propose that the following character sequence for
joke markers:

:-)

Read it sideways.  Actually, it is probably more
economical to mark things that are NOT jokes, given
current trends.  For this, use

:-(

The computer science bulletin board at Carnegie Mellon had spent days on escalating jokes about a mercury spill in an elevator, and some readers, missing the joke, took the hazard notices literally. Scott Fahlman's fix was a protocol proposal in miniature: problem, marker, usage note — read it sideways — and a wry extension for marking what is not a joke, given current trends.

The suggestion took. The sideways face spread across ARPANET sites, sprouted variants, crossed into every messaging medium since, and was eventually succeeded by pictographs standardized by an international consortium — a committee-run descendant of a bulletin-board joke. The original post, meanwhile, vanished: no one archived bboards, and for twenty years the founding text of emotional markup survived only in memory.

In 2002 Jeff Baird and colleagues excavated it from decades-old backup tapes, timestamp intact — a landmark act of digital archaeology that gives this artifact its unusual precision. The post's grammatical stumble in the first sentence survives too, unedited: history is drafted, not proofread.


symbolics.com

DNS LEDGER — 1985

The commercial internet's founding document is a ledger entry: one Lisp-machine company, registering a name in a namespace nobody was using yet.

   .com registrations, 1985

   15 March        symbolics.com
   24 April        bbn.com
   24 May          think.com
   11 July         mcc.com
   30 September    dec.com

   registrations under .com by 1992:  fewer
   than 15,000

   registrations under .com by 2010:  more
   than 80,000,000

The Domain Name System replaced a single, centrally edited file that had listed every computer on the network — a system that could not survive growth. When the general top-level domains opened in early 1985, registration was free, handled under government contract at SRI's Network Information Center, and wholly unhurried: the panel opposite lists the entire .com namespace of 1985, five research-computing companies deep.

First in the ledger, on 15 March, was Symbolics — maker of dedicated Lisp machines, the artisanal hardware of the AI laboratory world. The irony matured slowly: the commodity economics of the internet whose namespace Symbolics inaugurated would help extinguish the Lisp-machine business within a decade. The name outlived the company, and in 2009 was sold as a collectible — the ledger entry itself become the asset.

Nothing about the registration was experienced as an event; infrastructure firsts never are. But the five quiet lines of 1985 are the seed inventory of the largest namespace humanity has built, the address book that branding, speculation, and commerce would swell past eighty million entries. Every URL in the world ends in a descendant of this list.


TheProject.html

HYPERTEXT — 1991

The web's first page defined the web, linked to everything known about it, and promised universal access to a universe of documents — then nobody saved a copy.

World Wide Web

The WorldWideWeb (W3) is a wide-area hypermedia
information retrieval initiative aiming to give
universal access to a large universe of documents.

Everything there is online about W3 is linked
directly or indirectly to this document, including
an executive summary of the project, Mailing lists,
Policy, November's W3 news, Frequently Asked
Questions.

Tim Berners-Lee proposed a hypertext system for CERN in 1989 and by Christmas 1990 had built the whole stack himself on a NeXT workstation: browser-editor, server, and pages describing the project, served from info.cern.ch. The public found out on 6 August 1991, when he announced the WorldWideWeb project on a Usenet newsgroup — a date often misremembered as the first page's birthday.

The page itself was a plain directory of links under a definition — the sentence in the panel — with no image, no styling, and no sense of occasion. It was also the web's root: everything about the project linked directly or indirectly back to it. No one preserved it as it evolved; the earliest surviving copy is a 1992 snapshot kept by the W3C, restored to its original address by CERN in 2013.

Measured against what the web became, the definition is among the most completely fulfilled design sentences in engineering. The artifact's broken provenance is the joke history plays on ambition: the medium that would archive everything failed, first, to archive itself.


The Trojan Room coffee pot

128 × 128 GREYSCALE — 1993

The first camera on the web pointed at a coffee pot in Cambridge, because the researchers upstairs were tired of walking down to an empty one.

The image was only updated about three times a
minute, but that was fine because the pot filled
rather slowly, and it was only greyscale, which
was also fine, because so was the coffee.

                        -- Quentin Stafford-Fraser

In 1991, Quentin Stafford-Fraser and Paul Jardetzky aimed a spare camera at the shared filter machine in the Cambridge Computer Laboratory's Trojan Room and served a tiny greyscale image to colleagues' workstations, updated a few times a minute — local plumbing for a local grievance, described by its author with the deadpan sampled opposite.

In November 1993, once browsers could display images, Daniel Gordon and Martyn Johnson connected the feed to the web, and a departmental appliance became the first live camera in the medium's history. Millions visited. There was never anything to see — which turned out to be the discovery: liveness itself was the content, the fact of a real place breathing in real time through the screen.

The feed ran until 22 August 2001, when the Laboratory moved buildings and switched it off before the eyes of the world's press; the final pot fetched £3,350 at auction from a German news magazine. Every webcam, doorbell stream, and traffic camera online descends from this pot — the web's founding proof that people will watch anything, provided it is actually happening.


Merry Christmas

GSM SMS — 1992

The first text message was a season's greeting sent from a desktop computer to a phone that could not reply — fifteen characters into a channel built for bookkeeping.

Merry Christmas

SMS existed on paper for most of a decade before it existed in the air: drafted into the GSM standards in the mid-1980s, its message length fixed at 160 characters to fit the network's signaling channels — capacity reserved for the system's own housekeeping, where a short text could ride for free. Handsets that could compose text did not yet exist when the network was ready to carry it.

So the first SMS left from a computer: on 3 December 1992, Neil Papworth, a young engineer at Sema Group testing the message center his team had built for Vodafone, typed a greeting to Richard Jarvis of Vodafone, who read it on an Orbitel 901 at a company Christmas party. The handset could receive but not send; the medium's first message could not be answered in kind.

Texting became, for two decades, the most used data service on earth — trillions of messages a year, a compressed idiom shaped by the 160-character frame, and the ancestor constraint of microblogging. In 2021 the first message was sold, as these things now are, in tokenized form for charity: six figures for two words that had cost nothing.


just setting up my twttr

STATUS ID 20 — 2006

By 2006 a first transmission was no longer infrastructure but content: five lowercase words about the account that carried them.

just setting up my twttr

    -- @jack, 21 March 2006, status ID 20

On 21 March 2006, at a struggling podcast company in San Francisco, Jack Dorsey posted the prototype's first status update. The service was called twttr then — vowels were unfashionable, and the SMS shortcode thinking showed: its 140-character limit was carved from text messaging's 160, leaving room for a username. Colleagues posted the same phrase that day as their accounts initialized; Dorsey's, status ID 20, was first.

The post is pure self-reference — a message about joining the medium, composed with the ceremony of a shrug. That was the innovation. The status update assumed an audience of whoever cared, about nothing in particular, forever; and the form the platform standardized — short, public, timestamped — turned out to be the shape of news, politics, and celebrity for the following two decades.

In 2021 the first tweet was sold as a non-fungible token for $2.9 million; at attempted resale a year later, bids peaked under $300 — the era's tidiest lesson in what provenance is worth, and to whom. The artifact closes an arc: from scripture chosen for posterity to a test string nobody saved to five words about setting up the account. The medium became the message's only subject, and the wire went quiet behind it.


Full provenance for every piece is maintained in the SOURCE MATERIAL archive, one canonical record per artifact. Editorial line for no. 003: every panel is a verified quotation of the historical record — or, where the original is lost, the participant's documented recollection, labeled as such in the record and on the page. Nothing is paraphrased into the artifact's voice.

SOURCE MATERIAL is a recurring series about historically significant code and the systems it ran on. Each piece is fact-checked against primary sources where they survive; reproduction status (verbatim, abridged, reconstructed, compiled) is declared per artifact, and corrections are recorded in a versioned archive.

Set in Helvetica and DejaVu Sans Mono on US Letter. Body pages follow the series plate: title band, deck, specimen panel, three untitled prose blocks, and a footer of fragments. The palette is cream, ink, and one accent per issue; no. 003 wears signal blue.

curated, written, compiled, and typeset by Steven Delpercio · delpercio.dev