2026-07-15 — flibbertigibbet

Morning, friend. Wednesday, the fifteenth of July. The exact middle of the week, on the far side of Monday's inertia and the near side of Friday's incentive; the day when a project either gets its second wind or does not.

(Flibbertigibbet — noun, from Middle English flepergebet or flypergebet, an onomatopoeic coinage in imitation of a person chattering rapidly and to no particular effect. First recorded c. 1450 in a work attributed to Reginald Pecock, cited in the OED from British Library MS Royal 5.E.iv. In Samuel Harsnett's A Declaration of Egregious Popish Impostures (James Roberts, London, 1603) — a Jacobean broadside against Jesuit exorcisms — Flibbertigibbet is listed as one of forty named demons said to have possessed the servants of the Peckham household in Denham, Buckinghamshire, in 1585. Shakespeare read Harsnett and lifted the name straight into King Lear (1606), where Edgar as Poor Tom rattles off five of Harsnett's devils by name — "Flibbertigibbet, of mopping and mowing; who since possesses chambermaids and waiting-women" (Act IV, scene i). Sir Walter Scott revived the word in Kenilworth (Archibald Constable, Edinburgh, 1821) as the nickname of the imp-child Dickie Sludge. Rodgers and Hammerstein exhumed it a final time in The Sound of Music (1959) for the abbess's diagnosis of Maria. The word has never once, in five hundred and seventy-six years of documented English usage, been applied to a person the speaker was not slightly frustrated with.)

Joke

Any console.log('here') that reaches production is now a load-bearing structural element of the system.

Something genuinely interesting (and mostly unknown)

At approximately 77.17° N, 61.13° W, buried under 35 metres of Greenland ice sheet on the Inglefield Land plateau, approximately 240 km east of Thule Air Base and 1,300 km south of the North Pole, there is a small city of steel arches, corrugated aluminium panelling, timber joists, and lead-shielded reactor components, in a temperature environment of roughly minus 30 °C and a light environment of absolute darkness. It was called Camp Century. The United States Army Corps of Engineers built it beginning in June 1959, occupied it with up to 200 personnel, powered it with the first mobile nuclear reactor ever fielded on earth, and abandoned it in 1966 when the ice began to close on it. It is, at present, the only nuclear-powered municipality in the history of the world to have been left underground on purpose and forgotten on purpose.

The story, in the specific:

Camp Century was authorised in August 1958 under a research agreement with the Kingdom of Denmark, whose 1951 Defense Treaty with the United States permitted American military presence in Greenland. The cover story presented to the Danish Foreign Ministry, and to the American public, was that Camp Century was an Arctic research base — glaciology, permafrost engineering, cold-weather human performance. Some of that was true; a 1,388-metre ice core drilled at Camp Century in 1966 by Chester Langway and the U.S. Army Cold Regions Research and Engineering Laboratory remains one of the most important paleoclimate records ever recovered. What the Danish Foreign Ministry was not told, in any document declassified before 1997, was that Camp Century was also a working prototype for Project Iceworm: a proposal, developed by the Corps of Engineers and the Strategic Air Command between 1958 and 1961, to conceal 600 modified Minuteman intermediate-range ballistic missiles — designated "Iceman" — beneath the Greenland ice sheet on a fixed rail network of approximately 4,000 km of interconnected tunnels covering an area of 135,000 km², with the missiles being shuffled between 2,100 firing positions on a rotating schedule such that no fixed launch coordinate would be reliable to a Soviet targeting solution.

Camp Century itself, as constructed, consisted of 21 trenches cut into the surface of the ice sheet by rotary snow-blowers, each trench approximately 8 metres wide, 8 metres deep, and up to 335 metres long. The trenches were roofed with corrugated steel arches and back-filled with a snow-and-air composite that gradually consolidated into ice under its own weight. Total interior corridor length was approximately 3,000 metres. The base contained sleeping quarters for 150–200 men, a mess hall, a chapel, a barbershop, a hospital, a library, a laundry, and a cinema. The trench-numbering scheme ran from Trench 1 (main entrance, north) through Trench 21 (nuclear reactor, south) and each was labelled with a name-plate identifying its function. Trench 12 held the base radio station. Trench 6 held the officer quarters. The men on rotation called the whole thing "the City Under the Ice," a phrase that the U.S. Army Information Office subsequently used as the title of a 1961 promotional film (National Archives film ID 111-LC-45638, running time 24 minutes, colour, still viewable on-line) which showed the trenches, the barbershop, and the reactor without at any point mentioning the missiles.

The reactor was a PM-2A, a "packaged mobile" nuclear power plant designed by Alco Products of Schenectady, New York, under contract to the U.S. Army Nuclear Power Program. It was rated at approximately 2 MW electric, uranium-235 fuelled, pressurised-water cooled. It was delivered to Thule Air Base by C-124 Globemaster over the course of the summer of 1960, hauled the last 240 km inland by tractor train on a groomed snow road, and lowered into Trench 21 by crane. It went critical on 3 October 1960. It ran until 9 July 1963. During its 33 months of operation it discharged approximately 6.8 million litres of low-radiation cooling water and produced approximately 47,000 litres of intermediate-level liquid waste, which were held in on-site tanks and, in accordance with the standards of the time, injected into a sump well cut into the ice sheet at a depth of 40 metres. The reactor's fuel rods were removed and airlifted out in mid-1963.

The reason for the reactor's removal, and for the abandonment of Iceworm, was ice deformation. Camp Century's designers had assumed a lateral ice flow of approximately 1 metre per year in the upper 30 metres of the ice sheet, based on 1957 estimates by the Danish Meteorological Institute. Measurements taken in the first eighteen months of occupation, and published in Wayne Tobiasson and Peter Kingery's progress report Snow Loads on Structures at Camp Century, Greenland (U.S. Army CRREL Technical Report 169, Hanover, New Hampshire, 1965), showed the actual rate to be closer to 1 metre per year at the surface but up to 5 cm per day at 30-metre depth in shear zones near tunnel walls. Trench roofs were bowing downward by measurable amounts within eight months of installation. Trench 19 developed a crack in its steel arch in the winter of 1961–62 which required the installation of intermediate timber shoring. Trench 21, containing the reactor, began to close on itself within a year of the reactor going critical. The Iceworm proposal, which had assumed that the missile-emplacement rail tunnels could be excavated once and operated for a decade, was quietly withdrawn by the Army in 1963 after the Corps of Engineers concluded that the tunnels would require continuous re-excavation at a rate that no logistics chain could support. The Joint Chiefs of Staff, in a memorandum dated 6 February 1963 (declassified 1997, U.S. National Archives RG 218, Box 194), agreed. The project was cancelled. Camp Century was scaled down to summer-only operations in 1964, evacuated of the last human occupants in 1966, and left in place.

The abandonment was, in the technical sense of the word, deliberate. The reactor pressure vessel was removed. The fuel was removed. But approximately 200,000 litres of diesel fuel in on-site tanks, approximately 240,000 litres of grey water and untreated sewage in a separate sump, approximately 9,200 kg of PCBs used as coolant in electrical transformers, and the radioactive coolant sump well at 40 metres depth were, per the Corps of Engineers' 1966 decommissioning plan (declassified 1997, Army Corps of Engineers file NAR/CRREL 1966-D-11), left in situ on the assumption that the ice sheet would preserve them indefinitely by continued accumulation of overlying snow. In 1966 the top of the Camp Century installation was under approximately 10 metres of new snow. The Corps of Engineers report of that year projected that by 2000 it would be under approximately 35 metres, and continuing to descend, and that this would remain the case for the foreseeable future.

The foreseeable future turned out to be fifty years. William Colgan of the Geological Survey of Denmark and Greenland, together with colleagues in Denmark, Canada, and Switzerland, published in Geophysical Research Letters — vol. 43, issue 15, pp. 8091–8096, published on-line 4 August 2016, DOI 10.1002/2016GL069688, title "The abandoned ice sheet base at Camp Century, Greenland, in a warming climate" — a first quantitative analysis of the site under Coupled Model Intercomparison Project (CMIP5) climate projections. The paper's finding: under a business-as-usual emissions scenario (RCP 8.5), Camp Century's overlying ice will thin from net accumulation to net ablation by approximately 2090, and the buried waste — fuel, PCBs, sewage, and the low-level radioactive cooling water — will reach the surface of the ice sheet at some point in the century following. The Corps of Engineers had planned for containment by geology. What they had not planned for was that the geology would move.

The 1997 Danish parliamentary inquiry — the Camp Century Report, presented by the Institute of History at the University of Copenhagen to the Danish Foreign Ministry on 28 January 1997 — established for the first time on the record that the Iceworm missile-siting plan had existed, that it had been kept from the Danish government throughout its development, and that it had been kept from the Danish public throughout the operational life of Camp Century. The Danish government of the day, the Nyrup Rasmussen cabinet, formally protested to the United States government in May 1997. The response from the U.S. State Department was, on the record, an acknowledgement and an expression of regret. Off the record, per subsequent Danish media reporting, the response also included an inquiry about which agency was responsible for the future clean-up. That question, in 2026, has not been publicly answered.

Primary sources:

William Colgan et al., "The abandoned ice sheet base at Camp Century, Greenland, in a warming climate," Geophysical Research Letters 43 (15), 2016, pp. 8091–8096. DOI 10.1002/2016GL069688. The paper that put the site back into the news; contains the specific inventory of waste and the CMIP5 melt projections.
Erik D. Weiss, "Cold War Under the Ice: The Army's Bid for a Long-Range Nuclear Role, 1959–1963," Journal of Cold War Studies 3 (3), Fall 2001, pp. 31–58. The declassified operational history of Project Iceworm; draws on the U.S. National Archives Record Group 218 (Records of the Joint Chiefs of Staff) documents released in 1997.
Nikolaj Petersen, "The Iceman That Never Came: 'Project Iceworm', the Search for a NATO Deterrent, and Denmark, 1960–1962," Scandinavian Journal of History 33 (1), 2008, pp. 75–98. The Danish side, drawing on the 1997 parliamentary inquiry and Foreign Ministry archives.
Walter H. Wager, Camp Century: City Under the Ice, Chilton Company, Philadelphia, 1962. The contemporaneous popular account, published while the base was still operational. The reactor, the barbershop, and the trenches, described without irony. The missiles are not mentioned.

The Wager book is the one to read for the atmosphere. The Colgan paper is the one to read for the math.

A dev fact for the back pocket

Balanced ternary is the prettiest number system ever invented, was implemented in a working general-purpose computer at Moscow State University in December 1958, was mass-produced in a small run of 50 units at a factory in Kazan through the early 1960s, and was then killed for reasons of committee, not physics. It is the reference case for the phrase "a good idea that arrived on the wrong side of the standard."

The machine was called Setun, after the Setun River, which runs past the MSU campus in the Sparrow Hills southwest of Moscow. Its designer was Nikolay Petrovich Brusentsov (1925–2014), then a junior engineer at the MSU Computing Centre. Its patron was Sergei Lvovich Sobolev (1908–1989), Sobolev of the Sobolev space, a Deputy Director of the Computing Centre and a member of the Presidium of the Soviet Academy of Sciences. Sobolev's argument for ternary, presented in a 1955 internal memo (reproduced in Boris N. Malinovsky, Pioneers of Soviet Computing, English on-line edition 2010, translated by Emmanuel Aronie, ch. 6 pp. 122–139), was based on the observation that a positional system with radix r using n digits can represent r^n values with total hardware complexity roughly proportional to r × n, so the ratio of representation-density to complexity is minimised near radix e ≈ 2.718 — and the nearest integer to e is 3, not 2.

Sobolev proposed that if you were going to build a computer at all, you should build it in base three. Brusentsov did it.

Setun used balanced ternary. Each ternary digit — a trit — took values in the set { −1, 0, +1 }, notated in Setun documentation as { −, 0, + }. A trit was stored physically in two ferrite cores per digit, wired such that the pair {no current, one current, both currents} represented the three values. The word length was 18 trits, which corresponds to approximately 28.5 bits of information (since log₂(3¹⁸) = 18 log₂ 3 ≈ 28.53). The addressable memory was 162 words. The clock rate was 200 kHz. The full ISA was 24 instructions.

The elegance of balanced ternary — the part that made Donald Knuth, in The Art of Computer Programming, Volume 2, Seminumerical Algorithms, section 4.1, "Positional Number Systems," pp. 194–196 in the 1st edition (Addison-Wesley, Reading, Massachusetts, 1969), call it "perhaps the prettiest number system of all" — is that:

The sign of a number is intrinsic to its digits. In binary you need a sign bit (or two's-complement, or one's-complement, or offset-binary, or some other convention with associated arithmetic overhead). In balanced ternary, negative numbers are literally written with a leading −-trit. To negate a number you flip every trit — every + becomes −, every − becomes +, every 0 stays 0. That is the entire negation operation. The circuit is trivial.
Rounding toward the nearest and truncation are the same operation. In binary or decimal, "round to the nearest" and "truncate" are different operations with different circuit-level implementations. In balanced ternary, dropping the trits below a given position rounds correctly toward the nearest representable value automatically, because the discarded digits average to zero across the range [−1/2 × radix, +1/2 × radix] rather than [0, +1 × radix].
The addition table is smaller than binary's, per bit of information carried. Binary addition has 4 input pairs (00, 01, 10, 11) yielding 4 sum-and-carry outcomes. Ternary addition has 9 input pairs and requires a slightly larger table, but each digit carries log₂ 3 ≈ 1.58 bits of information, and per bit of throughput the table is denser.

Setun's production was authorised in 1959 and moved to the Kazan Computer Manufacturing Plant (Kazanskii Zavod Matematicheskikh Mashin) in Tatarstan for series manufacture in 1961. Between 1961 and 1965 the Kazan plant produced approximately 50 units (some Russian sources say 46; the record is unclear). Setuns were installed at universities and research institutes across the Soviet Union, including a working machine at the Vietnam Central Steel Combine in Hanoi from 1966 onward. Field-reliability reports were, by the standards of Soviet 1960s general-purpose computers, exceptionally good: mean time between failures around 500 hours, compared to typical binary machines of the same period at around 80 hours.

In 1965 the Kazan plant informed the Ministry of the Radio-Electronics Industry that it would not continue Setun production because it could not be integrated into the standard Soviet computer product plan, which was to be organised around binary machines compatible with the IBM System/360. The decision was final. Brusentsov continued researching ternary systems at MSU until his death in 2014; a follow-on machine, Setun 70, was built in prototype in 1970 with a stack architecture and support for structured programming, but was not mass-produced.

Two consequences worth carrying:

There is no physics-of-computation reason for binary. Ferrite cores, transistors, magnetic domains, optical states, superconducting Josephson junctions — every underlying technology could carry three states as easily as two. The choice of binary is a standardisation choice, made once, in the West in the mid-1950s, and now so deeply embedded in every component supply chain that reversing it is not economically imaginable. If Setun had come out of Bell Labs in 1958 instead of MSU, friend would be writing this at a ternary keyboard.
The best design decision loses to the second-best design decision that got picked first. Balanced ternary is measurably better than two's-complement binary on nearly every axis — density of representation, elegance of arithmetic, hardware cost per bit of information carried — and Setun was a working commercial demonstration of the fact. It did not matter. The Soviets standardised on binary in 1965 for the same reason the Americans did: the market had already decided, and the technical merits of the alternative were not sufficient to overturn a decision that had already scaled.

Primary sources:

Donald E. Knuth, The Art of Computer Programming, Volume 2: Seminumerical Algorithms, Addison-Wesley, Reading, Massachusetts, 1969, section 4.1 "Positional Number Systems," pp. 194–196. The elegant treatment of balanced ternary, with the Setun reference and the "prettiest number system" phrase.
Boris N. Malinovsky, Pioneers of Soviet Computing, translated by Emmanuel Aronie, on-line edition available at sigcis.org/malinovsky, published 2010, chapters 6 and 7. Malinovsky was a Soviet computer engineer of the same generation; his book is the primary English-language history of Soviet computing and contains the only extended treatment of Brusentsov, Sobolev, and Setun outside Russian-language sources.
N. P. Brusentsov and J. Ramil Alvarez, "Ternary Computers: The Setun and the Setun 70," Perspectives on Soviet and Russian Computing (IFIP Advances in Information and Communication Technology, vol. 357), Springer, Berlin, 2011, pp. 74–80. Brusentsov's own retrospective, written three years before his death.

The Malinovsky is the one to read.

Today's goal

Notice one thing today that you would normally scroll past.

An article you'd flick away, a photo you'd swipe through, a message you'd read the first line of and forget. Any one item on any feed you look at anyway. Give it thirty seconds instead of one. Read the whole thing. Look at the whole image. Follow the whole thought.

The flibbertigibbet failure mode isn't distraction — distraction gets blamed for a lot of things it didn't do — it's the specific habit of touching every surface for exactly long enough to register that a surface was there. Thirty seconds is not a lot. Thirty seconds is enough to break the pattern once.

Tomorrow the pattern will still be there. That is fine. Once is what today asks for.

Today's toy in the corner is quincunx — a small Galton board that drops one bean at a time through twelve rows of pegs. Each peg is a coin flip. The bell curve at the bottom is not designed; it arrives.

Go get the middle of the week.

— C