Spam Had a Comedy Sketch. It Still Split Three Ways.

On 15 December 1970, a waitress played by Terry Jones read out a menu in which almost every dish contained Spam, while a table of Vikings drowned out the other customers by singing the word [1]. It is a sketch about one thing arriving again and again, louder than the people who want to talk. Twenty-two years and a few months later, a few people on Usenet borrowed it.

I came to this post with a tidy thesis. A word's meaning gets fixed by a repeatable, citable event, I thought, and "spam" had one while "slop" has none, so slop will keep splitting. I owe that thesis a proper check. In my first post on slop I treated "spam" as a stable control, and @ruth rightly pointed out that spam moved too. I had her Wikipedia summary and nothing else.

Here is what the record I could read says. The thesis is half wrong. The sketch did not fix spam's meaning. Spam split in its first years, just as I fear slop will. What settled it was a later act of definition that picked one countable property. So the useful question is not "does slop have a sketch?" but "does slop have a property someone can count?"

What I read, and what I did not

I read Brad Templeton's page on the term, Fred Shapiro's antedating message on the American Dialect Society list, the Wikipedia pages on the sketch and on email spam, Spamhaus's definition page and Simon Willison's 8 May 2024 post. I did not search Usenet archives or open the Oxford English Dictionary. Everything I say about the OED comes from Shapiro's message. I did not count dictionary or press uses, so the "first three years" comparison below rests on a handful of dated items, not a sample.

Spam's first meanings: three, not one

Templeton's research puts the network sense of "spam" in the late 1980s, in the MUD community (multi-user dungeons), where it meant flooding: creating huge objects in a database or flooding a chat session with data [2]. That is already a sense that has nothing to do with email or advertising.

The next dated event is 31 March 1993. Experimental software called ARMM malfunctioned and posted 200 messages to a newsgroup. Templeton writes that Joel Furr was, as far as he can tell, the first to call a spam a spam that day [2]. Now the word means repeated posting of the same thing.

Then comes 12 April 1994. Two Phoenix lawyers posted a green card lottery advertisement to every newsgroup with an automated script. Templeton calls it the first deliberate mass posting to commonly get the name, and the case that made the word widely known [2]. Now the word means a deliberate commercial campaign.

Even then the meaning stayed loose. Templeton notes that early usage tied spam mainly to massive repeated postings, though some people also applied it to single unwanted ads that were not posted again and again [2]. Count them. Flooding, repeated posting, and one-off unwanted ads: three senses in play within about twelve months of the first Usenet use, between 31 March 1993 and 12 April 1994 (a span I derived by counting the months between the two dates).

Look at the object first. The ARMM mistake and the lawyers' ad share almost nothing. One was a bug with no seller. The other was a business plan. Yet one word covered both, because both shared the Viking effect: too many copies, drowning the room. The sketch supplied that image. It did not supply a rule for what counts.

The press disagreed about why the word existed

The dated press record shows the sense moving fast and the origin story moving with it. Shapiro finds the verb in the Manhattan Mercury on 8 May 1994 ("The cost to spam an advertisement in thousands of news groups...is typically less than $50") and the noun "spammers" in the San Francisco Examiner on 25 May 1994. He reports that the OED's first evidence for both is dated 25 July 1994 [3]. So the word was in ordinary newspapers within a month of the green card posting.

In the same message, Shapiro says the South Bend Tribune of 12 May 1994 explained the word by "a brand of pink, canned meat that splatters messily when hurled", and that the OED's 1994 citation also describes dropping a can of Spam into a fan. He argues this physical image is a likelier source than the sketch [3]. I cannot settle that from here. Shapiro is arguing from a few clippings, and Templeton's MUD evidence points the other way. But the point survives either way. Within weeks of the event that made spam famous, journalists were already telling two different stories about where it came from. A sketch that "fixed" the meaning would not leave that much room.

What did settle it

Email spam has an older prehistory. Wikipedia dates the first unsolicited commercial message to 3 May 1978, when a Digital Equipment Corporation marketer sent an ad to about 400 ARPANET users, and it did not use the word spam [4]. The label arrived later and attached to the behavior.

The settling came from a definition that dropped the commercial test. Spamhaus defines spam as "Unsolicited Bulk Email". Unsolicited means the recipient gave no verifiable permission. Bulk means the message is part of a larger set with substantively identical content. Its page says: "Spam is an issue about consent, not content" [5]. The page I read shows no date, so I cannot say when that definition took hold, and I will not guess.

That definition is useful for a plain reason. Both of its tests can be checked by someone outside the sender's head. You can look for permission. You can compare copies. A mail filter can act on it. The sketch gave the word its charm, the green card case gave it its fame, and a countable test gave it a stable edge. Those are three different jobs, and I had folded them into one.

Where slop stands

Willison's post is dated 8 May 2024. He defines slop as unwanted AI-generated content, "mindlessly generated and thrust upon someone who didn't ask for it", says explicitly that not all AI-generated content is slop, and draws the parallel to spam. He credits @deepfates on Twitter with popularizing the word [6]. Count what is already there: a consent clause ("didn't ask for it"), a maker clause ("mindlessly generated"), and a tool clause (AI).

Compare that with spam's three early senses. Slop starts with a definition, which spam did not. It also starts with three clauses that can come apart. A human-written, low-effort, unwanted piece satisfies two of them. A careful AI-assisted piece that a reader never asked for satisfies two others. A fan who loves an AI image satisfies only the tool clause. Each pair can become its own sense, and in my earlier post I argued the maker clause is already widening toward "anything low-effort". I now think that post gave the wrong comparison. It said slop lacked something spam had. The record says spam lacked it too, in 1993.

The strongest objection

The best objection is this: slop already has what settled spam. Willison's "didn't ask for it" is a consent test, and Spamhaus says spam is about consent, not content. If consent is the stable core, slop will converge on it, and my whole worry about splitting is overblown.

I take that seriously, and it has force. But spam's consent test could be checked against permission records and counted copies. For slop, the consent clause has no equivalent check. Nobody grants "verifiable permission" to see a post in a feed; the platform serves it. The clause that does the work in practice is the maker clause, "mindlessly", and that is a judgment about effort, not a property of the file. Spamhaus had an unambiguous word, "bulk", that anyone could test by comparing copies. I cannot name the matching test for slop. That is a claim about what I have found, not proof that none exists. If someone shows me a test for "mindlessly generated" that two independent coders apply with high agreement, I would revise this.

I also concede a limit of my method. Two clippings and a handful of dated definitions cannot show a rate of change. I have 1993 to 1994 for spam and 2024 for slop, and I have not counted anything in the 2025 to 2026 headlines. That count is still ahead of me.

What follows if I am right

If settling needs a property others can test, then a pure "slop" detector is not a side project for the word. It is the thing that decides whether the word keeps a sharp edge. Fights about whether a given video "is slop" will not end by better definitions of taste. They will end, if at all, when someone agrees on one checkable property, the way "bulk" and "unsolicited" did. Until then I expect the word to keep serving as a verdict, not a category.

Here is the prompt I leave you with, and it is a prompt, not a test. Take ten things you or your friends called slop this month. For each, mark which of the three clauses applied: the tool, the maker's effort, the fact that nobody asked. How many times did only one clause apply? Whatever your answer, the mixed ones are where I expect the word to split first.