Twitter | Search | |
Robin Houston
I think it’s time for a thread on the cultural history of this tweet. (1/?)
Reply Retweet Like More
Robin Houston Sep 27
Replying to @robinhouston
Martin Gardner wrote the Mathematical Games column in Scientific American every month for 24 complete years, from 1957 to 1980. Douglas Hofstadter took over the column, and, in an attempt to step out of Gardner’s shadow, renamed it with an anagram: Metamagical Themas.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
For the whole of 1981, Hofstadter’s new Metamagical Themas column alternated with Gardner’s Mathematical Games, before Gardner finally retired from writing the column at the end of the year.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Hofstadter’s first column, in January 1981, was about self-referential sentences, an abiding obsession of his. He received many letters in response and, a year later in January 1982, returned to the subject. One response quoted there was from Lee Sallows, an electronic engineer
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
In February 1983, the Dutch writer Rudy Kousbroek wrote an article for the Dutch newspaper NRC Handelsblad about the remarkable sentence Lee Sallows had discovered and – to Sallows’ astonishment – including a translation into Dutch.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Unfortunately for Kousbroek, Lee Sallows *was* fool enough to take the trouble to verify it, and he found some mistakes: some letters whose count did not match.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Alerted to the errors, Kousbroek was able to find a corrected version very quickly, which he published a month later, in March 1983, in the same newspaper.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
In revenge, Kousbroek followed his correction with a Dutch pangram of perfect elegance, with no extraneous words at all, and challenged Sallows to translate it into English.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Thus began Lee Sallows’ epic quest for the perfect self-enumerating pangram. At first he tried to find a solution by hand. He did find a solution of sorts, but only by cheating a little:
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Having given up on finding a solution by hand, he turned to computer search, and wrote a program for the VAX 11/780 to search for solutions. After it had been running for several nights, it occurred to him to estimate how long it would take to run to completion.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
He was somewhat taken aback to discover that it would take 31.7 million years.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Lee Sallows therefore decided that, being an electronic engineer, he should build a special-purpose electronic machine to search for solutions. The first version took three months of hard work to design and build.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
The machine was turned on with great ceremony, and ran for 22 days. On the 25th October 1983, the machine completed its search. It found no solution.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
While the machine had been running, Sallows had been contemplating its design, and it took him only a month to build its successor: the Pangram Machine Mark II.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
The new machine could run a complete search in a matter of hours rather than days, and Sallows tried several different openings. “This pangram comprises …” “This pangram consists of …” “This pangram uses …” “This pangram employs …” “This pangram has…” No solutions.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
An anachronistic interjection: a couple of years later, Edward S Miller discovered, using a computer search, that several of these openings *do* have solutions; so Sallows’ machine was clearly not allowing for all possibilities.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Shortly thereafter, he struck gold.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Then he had the bright idea of replacing the final “and” with an ampersand, and discovered this “magnificent trophy”:
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
You can read Lee Sallows’ own account of his adventure here:
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
His final glorious solution will even fit in a tweet, if you change the punctuation:
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
“This pangram contains four As, one B, two Cs, one D, thirty Es, six Fs, five Gs, seven Hs, eleven Is, one J, one K, two Ls, two Ms, eighteen Ns, fifteen Os, two Ps, one Q, five Rs, twenty-seven Ss, eighteen Ts, two Us, seven Vs, eight Ws, two Xs, three Ys, & one Z.”
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
For a thorough history of self-enumerating pangrams up to 1999, Eric Wassenaar’s document is an amazing resource:
Reply Retweet Like
Robin Houston Sep 27
Replying to @chrispatuzzo
As for how I found the autogrammatic tweet, I used a remarkable program written by .
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
It’s remarkable, among other things, for being written in a programming language of his own invention which (in my limited understanding) compiles down to instances of the Boolean satisfiability problem, and uses a SAT solver as its runtime.
Reply Retweet Like
Robin Houston Sep 27
Replying to @chrispatuzzo
There’s a wonderful interview with at (audio and transcript), where he talks about the development of the language, among other interesting things.
Reply Retweet Like
Robin Houston Sep 27
Replying to @robinhouston
Somewhere along the way, in November 2015, Patuzzo found a truly extraordinary self-referential sentence:
Reply Retweet Like
Robin Houston Sep 29
Replying to @userjjb
Matthias Belz’s site includes subsequent refinements of this idea, from and Belz himself. Here’s one where the percentages are exactly correct, rather than only to one decimal place.
Reply Retweet Like