I’m on a bus passing Worcester, Massachusetts, partway between Cambridge and New York. Out the window is a wall of trees between which tiny New England towns occasionally peek. I’m listening to In Conflict by Owen Pallett. A group of researchers in Edinburgh grew a fully-functional human Thymus in a mouse. The Thymus produces T-cells, which seem important, but I don’t really understand anything about biology. It’s Monday evening.
Do you remember Single-Serving Sites? What Color Is the Empire State Buidling?, Is Lost a Repeat?, Instant Rimshot, The Abe Vigoda Status Page. Kottke has a list, many of which are now dead links. Or, if not properly dead, rendered irrelevant by time.
Single-Serving Sites were a strange transitional phenomenon. They often provided a packaged answer to a simple question about the off-net world: Is the actor Abe Vigoda alive? What color are the lights on the Empire State Building now? Is today Christmas? Or else they gave immediate access to some short piece of media that might frequently prove useful in online conversations. A rimshot, a sad trombone, William Shatner shouting “Khan”. Part of the humor came from giving a humble concept an entire URL to itself, which had a kind of irony at a time when most people thought of websites as large edifices created for adult purposes. It was funny that the domain system put isitchristmas.com on the same footing as amazon.com.
These sites were also part of a phase in the democratization of web production where things like domain registration, hosting, and screen scraping fell in cost and complexity, admitting new casual uses. You could make a single-serving site in an hour for about ten bucks.
Making websites trivial to make made people make different things.
My Single-Serving Sites
That Kottke post is from 2008, right around the peak of Single-Serving Sites by my reckoning. I made a number around that time. largeheartedgoat.com documented the music blog Largehearted Boy’s running obsession with the band The Mountain Goats (here’s my post announcing the site from 2006; one of my server moves since then seems to have killed it since then).
Another of my sites, The NY Times Explains the Ratings, documented that paper’s decision to include an often snarky one-liner with each movie review to explain the MPAA’s judgment. A typical example:
‘Music and Lyrics’ is rated PG-13 (Parents strongly cautioned). It has some sexual situations and naughty language. Parents of a certain age who see it with their children may have to endure some uncomfortable questions about the ’80s.
I found it funny to reduce the movie down to a summary of what about it might offend. I made that site in 2007. It’s also dead. The image above is courtesy of the Wayback Machine
Looking back now, this whole genre feel like an artifact of a different era online. While some single-serving sites may have included ads to offset hosting costs, they weren’t business or services aimed at making a profit. They were also not monetizable content for someone else’s social network, not part of the exhausting contemporary economy Ian Bogost has called hyperemployment. They were gloriously pointless, done for their own sake, for the few hour’s enjoyment in making them and little more.
However they also have some interesting properties that presage much of what the tech industry went on to do in the following years. Many of them embody pieces of information in single-use APIs that could be queried by programmatically. They’re like folk culture versions of the open inter-operable APIs that were a cornerstone of web startup culture before smartphones.
Further, single-serving sites oftentimes encapsulate information about things from the physical world like the Empire State Building’s lights, whether or not the New York Subway System’s L train is fucked, etc. This anticipated many of the current ideas around the Internet of Things and the drive to extend the web’s philosophy of unique identifiers into the built environment.
The Era of Twitter Bots
The rise of social networks, of Twitter and Facebook, killed single-serving sites. Social networks don’t technically disable single-serving sites or prevent people from making them in anyway. But they do change the semantics of the web, the nouns and verbs out of which we make meaning online.
The era of social media introduced a new unique identifier alongside domains: user accounts on social networks. The same frisson found in single-serving sites by putting istheltrainfucked.com at the same level as google.com now exists by giving a small idea a Twitter account and letting it speak within the network as if it was a peer of real individual people.
Hence Twitter bots have sidled into the role previously occupied by Single-Serving Sites. Take @samuelpepys, a bot created by Phil Gyford that tweets its way through 17th century British Parliamentarian Samuel Pepys diary or @everyword by Allison Parrish, which tweeted every word in the English language between 2007 and 2014.
These bots have the single-minded focus that’s familiar from Single-Serving Sites, but they also have some interesting new properties. They make spectacular use of time, slowly playing out systemic processes or drawing parallels between the past and the present.
Other bots take advantage of the real-time nature of Twitter by algorithmically responding to their informational surroundings. For example, Darius Kazemi’s @twoheadlines creates a hybrid headline by melding together two current news stories to hilarious effect (details on the algorithm):
Other bots make creative use of the fact that they live intimately amongst us. Rob Dubbin’s @oliviataters is a Markov chain teenager that seems to update the text from which it is generated based on a rolling list of tweets by real teenagers. Olivia almost always follows back users that follow her as well as responding to replies in ways that are shockingly like real teenage conversation even in ways that are troubling:
(I highly recommend the TLDR episode about Olivia: Olivia Taters, Robot Teenager)
A more abstract interactive example is @tweet2form, a bot that acts as a frontend to the Grasshopper algorithmic modeling environment, letting users create little 3D renderings from short lists of commands.
Because we live in closer proximity to them, people have a more intimate relationship to Twitter bots than they ever did to single-serving sites. For evidence of this, you need look no further than the outrage that ensued when @horse_ebooks was unmasked as being the self-promotional work of artists instead of an “honest bot”.
Think about that again for a second. People were disappointed that @horse_ebooks was an art project instead of the algorithmic spew of a spam bot. This speaks to another aspect of the aesthetic thrill of Twitter bots. They operate within an environment in which non-human operators seem to outnumber human ones. How many of the tweets you see pop out of a large site’s content management system or are scheduled in advance as part of a marketing campaign or are spam bots replying to you because of some keyword search you triggered?
This is the background radiation against which we conduct our social communications. We’re naturally fascinated by objects whose location is ambiguous between figure and ground. And when it seems to “naturally” produce something fascinating that naturalness itself makes it all the more miraculous. And there is some seriously fascinating unexplained stuff out there. I dare you to read this post about incomprehnsible tweets that may be spam bots or may be spies or drug dealers signaling each other without interest:
My Twitter Bots
I started making Twitter bots earlier this summer. I’ve become a little obsessed, making four already in the last month.
The first I produced was @uncannyxbot, a bot that generates summaries of non-existent X-Men comics:
Even though it was my first bot, I’m not sure I’ll ever beat @uncannyxbot for pure laughs. There’s something about the totally ridiculous, but predictable and repetitive, nature of the Marvel universe that seems to reliably produce hilarity when mashed up through a Markov chain.
In fact the bot was inspired by Rachel and Miles Explain the X-Men a highly entertaining podcast that does its best to make sense out of X-Men continuity. I was struck by just how often Rachel and Miles’ accurate explanations already sounded like Markov-generated gibberish.
My next bot arose from a discussion at this year’s Boston Game Loop about algorithmically generating rules of games. I found a site that catalogs the rules of card games and @rulesofcards was born. Here’s a recent example:
Its tweets tend to end up tantalizingly close to comprehensible. You could almost imagine following them in some real, if quite arcane, card game.
I made my third bot, @haskell_ebooks, to win an argument. I had a brief debate on twitter with a couple of friends about obscurantism in programming language discourse. I was arguing that the community around the Haskell programming language used particularly impenetrable language in its explanations. The conversation started with me joking that all Haskell blog posts already sounded like the output from Markov chains. So, naturally, I made that bot real to win the point.
I must have done something right because someone recently suggested using @haskell_ebooks tweets in programmer job interviews.
My most recent bot is more personal and I’m correspondingly more embarrassed of it. @lifetoll makes hyperbolic comparisons between mundane personal difficulties and real natural disasters. It’s meant to straddle the line between self-indulgent and self-mocking. It tends to come-off extremely emo:
Lately, I’ve been adding new ever-smaller life challenges to it, trying to bring out the irony of the juxtaposition:
Like all Twitter bots. It’s small and weird and personal and a little bit alive.