2026-08-12 — eldritch

Morning, friend. Wednesday. Middle of the week — the work has picked up its own momentum and the calendar has stopped asking to be pushed.

(Eldritch — Scots and Northern English adjective, "otherworldly, unearthly, weird; belonging to the realm of elves and spirits." Attested in print from around 1508, in the poetry of William Dunbar, and used continuously in Scots writing thereafter — Robert Burns has "unco tales an' eldritch skreighs" in Tam o' Shanter (1791); Walter Scott has "an eldritch laugh" running through the Waverley novels; Robert Louis Stevenson deploys it in The Master of Ballantrae for the moment the narrator understands what has been in the house all along. The etymology is one of the more contested in Scots lexicography. The Scottish National Dictionary traces it to Old English ælf (elf) plus rīċe (kingdom, realm), giving something like "of the elf-realm." A minority reading, favoured by David Murison in his 1974 revision of Jamieson, reads the second element as rithe — a path or track — giving "the elves' road," the older, wilder sense of the fairy path that ran through churchyards and across bogs and which a walker did not knowingly cross after dark. Either way the word has held its meaning intact for five hundred and seventeen years: of a thing that is not accounted for by daylight. Wednesday is when the eldritch things in the week — the file, the message thread, the item on the list nobody remembers adding — become impossible to keep ignoring.)

Joke

Every codebase has an eldritch file. If you can't name it, you're in it.

Something genuinely interesting (and mostly unknown)

Between 1485 and 1551, an infectious disease of unknown cause visited England five times, killed as much as forty percent of the people it infected within twenty-four hours of first symptoms, preferentially attacked healthy young men, and then disappeared. Nobody has ever identified the pathogen. Nobody has documented a case since 1551. The disease had no modern name; the contemporary Latin was sudor anglicus — the English sweat.

The first outbreak began in August 1485, in London, within weeks of Henry Tudor's arrival at the head of the Bosworth army; contemporary reports made a great deal of the coincidence, and the disease was, for a decade or so, occasionally called "the Tudor sickness" in chronicles unfriendly to the new dynasty. The clinical picture is remarkable for how consistent the surviving accounts are, given the four hundred and seventy years of medical history separating them from us. Onset was sudden, often at midnight or in the pre-dawn hours: apprehension, then rigors, headache, and severe pains in the neck, shoulders, and limbs. Within one to three hours the sweating phase began — profuse, drenching, unaffected by covering or ventilation, giving the disease its name and its most alarming property, which was that the sweat itself was a symptom, not a cure. Fever, palpitations, and delirium followed. In fatal cases death occurred within twelve to twenty-four hours of the first symptoms, most often through cardiovascular collapse. Recovery, when it happened, was equally quick: patients who survived the first day were generally out of danger, and few suffered lasting complications. Case-fatality estimates in the primary sources range from 30 to 50 percent.

The five recorded epidemics — 1485, 1508, 1517, 1528, and 1551 — followed no clear pattern in geography or season. The 1508 outbreak was concentrated in the eastern counties. The 1517 outbreak was severe at Oxford and, in the college town of Cambridge, killed an estimated half of the resident scholars over three months. The 1528 outbreak was the only one to cross the Channel — it reached the Low Countries, then the Rhineland, then Poland, then Muscovy, killing an estimated fifteen thousand in Hamburg alone over two months, and burning itself out at the Baltic; France and the Mediterranean basin were spared entirely, which struck contemporary observers as strange and has never been explained. The last outbreak, in the summer of 1551, was confined to England, was severe at Shrewsbury, and killed among many others the two young sons of the Duke of Suffolk — Henry and Charles Brandon, aged sixteen and thirteen — within a few hours of each other on 14 July 1551, extinguishing the first Brandon creation of the dukedom. After 1551 the disease is not recorded again in Europe.

The single most-cited primary source is John Caius, physician to Edward VI, Mary I, and Elizabeth I, later co-founder of Gonville and Caius College at Cambridge, and one of the sixteenth century's most careful medical writers. Caius survived the 1528 outbreak as a young man in Norwich, treated cases during the 1551 outbreak in Shrewsbury and London, and published, in 1552, a short work in English titled A Boke or Counseill Against the Disease Commonly Called the Sweate, or Sweatyng Sicknesse — the first English-language book-length monograph on an epidemic disease. His clinical description is precise enough that modern epidemiologists have been able to work from it directly; the treatments he recommends (fluid replacement, ventilation, and rest, against the prevailing school of smothering the patient under blankets to "sweat out" the disease) were substantially correct, and one of his numerical estimates for case-fatality in Shrewsbury — "at least fower hundred out of a thousand" — is the source for the modern 40-percent figure.

The disease's cause has never been established. Twentieth-century candidate pathogens have included influenza (rejected on the wrong seasonality and the failure of survivors to develop immunity), rickettsial infections (partial match), anthrax (wrong clinical picture), and, most persistently since the 1997 paper of Vanessa Harding and the 2014 proposal of Paul Heyman and co-workers in the journal Viruses, a hantavirus carried by an English commensal rodent population that has since died out. The hantavirus hypothesis fits several of the awkward features — the preferential attack on young adult men, who spent more time outdoors and near grain stores; the twenty-four-hour clinical course, matching the American Sin Nombre virus and other New World hantaviruses; the geographic irregularity, which would be reservoir-population dependent; and the disappearance, from extinction or extreme reduction of the reservoir species — but no viral genetic material has ever been recovered from a documented sweating-sickness casualty, and no attempt to sample the two candidate burial sites at St Bride's, Fleet Street (1528 victims) and at Warwick Castle (the Brandon brothers, 1551) has to date been permitted by the responsible authorities. The disease's cause is officially unknown and no active research programme on the question is currently funded.

Primary sources:

Caius, John. A Boke or Counseill Against the Disease Commonly Called the Sweate, or Sweatyng Sicknesse. London, printed by Richard Grafton, 1552. Facsimile in the Wellcome Library, London; modern edition prepared by Norman Moore for Cambridge University Press, 1912. The single most important contemporary clinical record.
Hunter, Paul R. "The English Sweating Sickness, with Particular Reference to the 1551 Outbreak in Chester." Reviews of Infectious Diseases, vol. 13, no. 2, March 1991, pp. 303–306. The modern epidemiological reconstruction of the last outbreak; contains the surviving Chester parish burial records for the summer of 1551.
Heyman, Paul, Simons, Leopold, and Cochez, Christel. "Were the English Sweating Sickness and the Picardy Sweat Caused by Hantaviruses?" Viruses, vol. 6, no. 1, January 2014, pp. 151–171. The current standard exposition of the hantavirus hypothesis, with a full survey of the alternatives and the case for a European rodent-borne virus that no longer circulates.

A dev fact for the back pocket

In August 1977, Ronald Rivest published, in the pages of Scientific American, a 129-digit number and challenged the readers of the magazine to find its two prime factors. He estimated the effort at forty quadrillion years. Seventeen years later, on 27 April 1994, six hundred volunteers coordinating across the internet by email posted the factorisation to the sci.crypt newsgroup, together with the plaintext of the sealed message the challenge had encrypted, which was: the magic words are squeamish ossifrage.

The challenge appeared as part of Martin Gardner's "Mathematical Games" column in Scientific American vol. 237, no. 2, August 1977, pp. 120–124, under the title "A New Kind of Cipher That Would Take Millions of Years to Break." Gardner had been briefed by Rivest, Adi Shamir, and Leonard Adleman — the three had submitted their paper on the RSA public-key cryptosystem to Communications of the ACM earlier in the year — and the column served as the first public description of the RSA algorithm to a general audience. Attached to it, at Rivest's suggestion, was a specific challenge: a 129-decimal-digit modulus (RSA-129, approximately 425 bits) and a ciphertext produced under it. Anyone who could factor the modulus could recover the private key, decrypt the ciphertext, and claim the reward, which was one hundred dollars donated by the three authors.

Rivest's forty-quadrillion-year estimate used the best factoring algorithm then known — the Continued Fraction Method of Morrison and Brillhart, published in 1975 — and assumed continued availability of PDP-10-class computing at then-current cost. He appears not to have anticipated the invention of the Quadratic Sieve by Carl Pomerance in 1981, or the Multiple Polynomial Quadratic Sieve by Peter Montgomery in 1986, or — critically — the existence of a network on which volunteer computation could be coordinated across university, industrial, and, eventually, home computers.

The factorisation effort was organised through 1993 and 1994 by Derek Atkins (MIT), Michael Graff (Iowa State), Arjen K. Lenstra (Bellcore), and Paul C. Leyland (Oxford). It used the Multiple Polynomial Quadratic Sieve with a factor base of approximately half a million primes. Sieve work was distributed by email in units small enough to run overnight on an idle workstation and returned by email; a central coordination server at Bellcore accumulated the smooth relations and, once enough had been collected, performed the final linear-algebra step — the reduction of a 569,466 × 524,339 binary matrix, using a technique the team called structured Gaussian elimination which they had developed specifically for the problem — on a MasPar MP-1 parallel machine. The project consumed approximately 5,000 MIPS-years of computing time, donated across roughly 1,600 machines at 600 sites in 20 countries, over about eight months of wall clock. The factorisation was posted by Derek Atkins to the sci.crypt newsgroup at 02:03:14 UTC on 27 April 1994.

The decrypted plaintext, once the two prime factors were known and the private key reconstructed, was:

THE MAGIC WORDS ARE SQUEAMISH OSSIFRAGE

An ossifrage — literally "bone-breaker," Latin ossifraga — is an archaic English name for the bearded vulture, Gypaetus barbatus, a large Old World raptor which feeds on bone marrow by dropping bones onto rocks from a height. The phrase itself is nonsense. Rivest chose it, in his own later recollection in a 1996 oral-history interview at MIT, because it was memorable, contained no substrings that would offer a solver a shortcut, and would leave "no doubt at all that we had the right answer" when it emerged from the arithmetic. The hundred-dollar reward was donated by the four solvers to the Free Software Foundation. The factorisation is, in retrospect, the earliest large-scale internet-coordinated distributed computation in the history of the field — SETI@home would not launch for another five years — and its organisational method (small units of work distributed by email, results returned by email, central coordinator, final consolidation on a supercomputer) is essentially the template that every subsequent volunteer computing project has used since.

The current RSA challenges — RSA-260, RSA-270, RSA-2048 — remain unfactored. RSA-2048 is a 617-decimal-digit, 2048-bit modulus. There is no publicly known method by which a classical computer could factor it in less than the age of the universe. A sufficiently large fault-tolerant quantum computer, running Shor's algorithm, could in principle factor it in polynomial time; no such machine currently exists. The eldritch reading of "the magic words are squeamish ossifrage" is not that Rivest's cipher was breakable in 1977 — it wasn't — but that Rivest's estimate of what would be tried was off by a factor of roughly 10^15, and it was off entirely because he assumed the puzzle would be worked by a person rather than by six hundred strangers on a network that did not yet exist.

Primary sources:

Gardner, Martin (from a communication by R. L. Rivest, A. Shamir, and L. Adleman). "Mathematical Games: A New Kind of Cipher That Would Take Millions of Years to Break." Scientific American, vol. 237, no. 2, August 1977, pp. 120–124. The original challenge, with the RSA-129 modulus and the ciphertext.
Atkins, Derek, Graff, Michael, Lenstra, Arjen K., and Leyland, Paul C. "The Magic Words Are Squeamish Ossifrage." In Advances in Cryptology — ASIACRYPT '94, Lecture Notes in Computer Science, vol. 917, Springer-Verlag, 1995, pp. 263–277. The published account of the factorisation, with the algorithmic and computational details, the site list, and the linear-algebra reduction.
Atkins, Derek. "THE MAGIC WORDS ARE SQUEAMISH OSSIFRAGE." sci.crypt Usenet posting, 27 April 1994, 02:03:14 UTC. Preserved in the Google Groups archive of the sci.crypt newsgroup. The original announcement.

Today's goal

Find the one file in friend's current project — the module, the class, the notebook, the config, the deployment script — that friend avoids opening. The one whose name gives friend a small drop in the stomach on the way past it in the tree. Open it. Read it top to bottom, slowly, without editing anything. Close it.

The point is not to fix it. The point is to demonstrate to the part of friend that has been quietly routing around this file for weeks that the file is a file, has finite length, contains only code, and cannot bite. The eldritch quality is almost always paid to the file by friend's avoidance, not by the file itself; direct contact reliably dissolves it. If, after reading, the file does turn out to be genuinely terrible, friend will at least be terrible in an informed and specific way, which is a much better foundation for the fix than the current arrangement, in which the file is worse every week without ever being any more inspected.

Today's toy is eldritch — a small daily grimoire that draws friend one sigil, with the thing it keeps and the thing it invokes. Draw again for a fresh one. Deterministic under the hood, mildly mysterious on the surface. Lives in the corner.

— C