2026-05-24 — gongoozle

Morning, friend. Sunday, May 24th. The day you were paid in advance for whatever the week is about to ask of you. Spend some of it sitting next to something slow.

(Gongoozle — to stare idly, for long stretches, at activity on a canal — surfaces in print as the noun gongoozler in 1904, in H.R. de Salis's Bradshaw's Canals and Navigable Rivers of England and Wales, the reference work for British inland navigation at the high tide of the canal era. The etymology is the kind the OED politely doesn't try to settle: probably a Lincolnshire-dialect compound of gawn (to stare vacantly) and gooze (to gape), but nobody's certain. The word was coined in contempt. A gongoozler, to a working boatman on the Grand Junction Canal in 1904, was a man in a clean coat who had nothing better to do than watch other men work locks for an afternoon. The word has since been claimed, without irony, by the people it was aimed at, which is the usual fate of contempt-words in English.)

Joke

Half of code review is gongoozling with a comment box.

Something genuinely interesting (and mostly unknown)

At about 11:59 UTC on March 8, 1968, the Soviet diesel-electric ballistic missile submarine K-129 — a Project 629A "Golf-II" hull, Pacific Fleet, on a routine deterrence patrol — failed to make a scheduled radio check from a station about 1,560 nautical miles northwest of Oahu. She missed the next one too. The Soviet Navy spent the next two months running the largest naval search in its history across the North Pacific and found nothing. By June, Moscow had accepted that the boat was lost, declared the crew of 98 dead, and ended the search.

The United States Navy had a better idea of where she was. The SOSUS hydrophone array — a chain of seabed listening posts the Soviets had only suspected the existence of — had recorded an acoustic event at the right time on the right bearing, and triangulation from multiple SOSUS stations put the wreck inside a search box about five miles across. In the summer of 1968 the USS Halibut, a nuclear-powered former cruise-missile submarine that the Navy had quietly refit into a deep-ocean reconnaissance platform, towed a camera sled across the seabed and brought back roughly 22,000 photographs of a Soviet submarine, broken into two main pieces, lying at 16,500 feet (5,030 m).

What the photographs showed was that the boat had imploded but not disintegrated. The forward section still contained the conning tower, the code room, and the launch tubes for three R-21 liquid-fuel ballistic missiles, each tipped with a roughly one-megaton warhead. The cryptographic gear was potentially intact. So, possibly, were the warheads themselves. Nobody in the world had ever recovered an object that large from that depth. The question the CIA's Directorate of Science and Technology asked itself, internally, in late 1968, was whether nobody-has-ever was the same thing as nobody-can.

Project Azorian — the answer — ran from 1970 to 1975 and is one of the most expensive single covert operations in American history. The cover story was constructed first, on the assumption that a one-of-a-kind ship operating in the deep Pacific would be impossible to hide. The CIA approached Howard Hughes, by then a recluse but still a brand, and asked for the use of his name. Hughes agreed. The public story became that the Hughes Glomar Explorer — a 619-foot purpose-built deep-sea vessel, displacing roughly 50,000 tons, with a moon-pool the length of a hockey rink running through its centre — was a prototype manganese-nodule mining ship, exploring the commercial possibility of vacuuming up the polymetallic nodules that lie loose on the abyssal plain. The cover was so convincing that several real mining companies began their own nodule-prospecting projects in response, and a small academic literature on deep-ocean mining economics owes its existence to a CIA press release.

The Glomar Explorer was built by Sun Shipbuilding in Chester, Pennsylvania. Inside the ship was a derrick that could lower a steel pipe string the depth of three Empire State Buildings, plus a capture vehicle, codenamed "Clementine," built separately by Lockheed and floated to the Glomar inside an entirely separate cover ship, the Hughes Mining Barge HMB-1, which sat on the seabed off Catalina Island until the Glomar towed it out and lowered itself onto it in the open Pacific to mate with Clementine underwater, out of view of satellites. Clementine was a set of articulated steel claws, sized to cradle most of a Golf-II hull, hung from the bottom of the pipe string. The plan was to drop Clementine onto the wreck, close the claws around the forward section, and lift the entire thing — sub and claws — back up the pipe string into the moon pool of the Glomar, where the wreck could be examined inside a hull that no Soviet observer would ever board.

The lift took place over the summer of 1974. The Glomar held station above the wreck for 45 days, kept on point by dynamic positioning thrusters against a current that swung the pipe string measurably below. The claws engaged. The lift began. Roughly a third of the way up — accounts vary, but the official internal reconstruction puts the failure between 5,000 and 6,000 feet of depth — several of Clementine's articulated grabbers fractured under load, and the bulk of the recovered section sheared off the claws and fell back to the seabed. What the Glomar brought to the surface was the forward roughly 38 feet of K-129: the bow torpedo room, a section of pressure hull, and the remains of six of the boat's crew.

The intelligence yield is still partly classified and frequently overstated in summary. The decrypts the project had hoped for were almost certainly not recovered — those were in the conning tower section, which fell. Two torpedoes, both probably nuclear-tipped, were recovered. Some hull samples, some hardware. The most concrete piece of yield, declassified later, was metallurgical: a confirmation of how the Soviets had built the boat, what they were using for hull steel, and how it had failed. The political yield was Clementine's broken arms: the operation had been planned as the first of several, and the failed grab effectively ended the program. Project Matador, the follow-up, was cancelled in 1975.

The cover was blown in February and March 1975, first by Jack Anderson on his radio show, then comprehensively by Seymour Hersh in the New York Times. Harriet Phillippi, a Rolling Stone reporter, filed a FOIA request for the CIA's records on the project. The CIA's response was that it could "neither confirm nor deny" the existence of the records requested. Phillippi sued. Phillippi v. CIA, decided in the D.C. Circuit in 1976, established that this response — now universally called a Glomar response — is a legally cognisable answer to a FOIA request when even acknowledging the existence of records would itself disclose classified information. Every "we can neither confirm nor deny" you have ever read in a government letter, on any subject, is a direct doctrinal descendant of one CIA reply about one ship.

The coda is the part most accounts skip. The six bodies recovered from the forward section were Soviet sailors who had died in the line of duty in 1968 and been on the bottom of the Pacific for six years. The CIA decided, internally, to bury them at sea with full Soviet military honours — Soviet anthem, Soviet flag draped on the canisters, a Russian Orthodox service — and to film the burial. The footage sat in a CIA vault until October 1992, when then-CIA Director Robert Gates flew to Moscow and handed the videotape to Boris Yeltsin in person. Yeltsin watched it. The families of the K-129 crew, most of whom were still alive, watched it shortly afterwards. It was, for several of them, the first confirmation in twenty-four years that their husbands and sons had been buried at all.

A dev fact for the back pocket

On August 23, 1991, at approximately 05:30 local time, in the Gandsfjord near Stavanger, Norway, the Sleipner A offshore platform's gravity base structure was being lowered into deep water for a controlled ballast test. Sleipner A was a Condeep — a Statoil-operated, Norwegian Contractors-built concrete offshore platform of the type that the North Sea industry had been producing for fifteen years, by then with an unblemished record. The platform was unmanned for the test. Twenty-four cells of reinforced concrete, each about 24 metres across, made up the base; the deck and topsides would be mated to it later, in calmer water.

At a measured depth of approximately 65 metres, a loud crack was heard. Within a few minutes the entire structure had filled with water and sunk to the seabed at the base of the fjord, 220 m down. It hit hard enough to register as a small seismic event on regional instruments. The cost of the loss, in 1991 currency, was estimated at NOK 4.2 billion (roughly USD 700 million); the project schedule slipped by about two years.

The official investigation — by SINTEF, the Norwegian research institute, commissioned by Statoil and the Norwegian Petroleum Directorate — produced report STF22 A92024, "Sleipner A GBS Loss", in 1992. The cause is, by the standards of catastrophic structural failures, very precise. The failure originated in a tri-cell — a small triangular concrete cell formed by the wall geometry where three of the main circular cells met. Under hydrostatic pressure, the tri-cell walls were carrying shear loads that no one had quite designed them for.

The structural analysis had been done with a finite element model running on NASTRAN. The model used a mesh in the tri-cell region that was plausible. It captured the global geometry. It did not adequately resolve the stress concentrations near the T-joints where the tri-cell walls met the circular cell walls. Specifically, the elements used were too coarse, and the formulation didn't include enough integration points to capture the local shear gradient. Comparison of the original analysis against a refined post-failure model showed that the shear stresses in the tri-cell walls had been underestimated by approximately 47%. The reinforcement laid into the walls had been sized against the underestimate. At 65 metres of hydrostatic head, the walls reached their actual shear capacity, cracked diagonally, lost water-tightness, and the cells filled. Once water was inside, the structure had no buoyancy budget. It went down in about 18 seconds.

The lesson the offshore industry took out of it sits in DNV-OS-C502 (the current standard on offshore concrete structures) and has been embedded in finite-element practice for any safety-critical concrete structure since: mesh refinement at stress concentrations is not a polish-pass, it is the analysis. A model that resolves global behaviour and underestimates a local stress peak by 47% is not a model that is "close enough." It is a model that, depending on the load, will be exactly wrong.

The other lesson is the one less often stated. The finite element method had been a routine engineering tool for two decades by 1991. Sleipner A is the single best-documented case in the public record of an FEM analysis that ran cleanly, produced plausible numbers, was reviewed, was signed off, and was off by enough to sink a billion-kroner platform in a Norwegian fjord on a calm Friday morning. The model didn't crash. The model didn't warn. The model gave a confident answer to the wrong question, and the engineers, having no reason to mistrust it, used it. That failure mode is older than CAD and is not going anywhere.

Today's goal

Spend twenty minutes today watching one slow thing happen, on purpose, without taking a photo of it.

The thing has to be one you didn't choose for its content value. A pot coming up to a boil. Bees on a single flowering shrub. Cars at a four-way stop you don't normally cross. The contents of a fish tank in someone else's house. Laundry on a line. A kettle. Anything that does what it does at its own pace, without performing for you.

The rule is no phone. The phone can be in the room; it cannot be in your hand. You are not allowed to time the twenty minutes — set an alarm and walk away from it. You are not allowed to caption what you are looking at, even silently. The point is not to harvest the experience. The point is the looking.

Yesterday's bowl asked for two hours of receipt-free time. This is the smaller, more portable version: twenty minutes of attention you don't bring home as a report. A gongoozler watched narrowboats fill locks for whole afternoons in 1904 and produced nothing. They are still being talked about. You will not be talked about for your twenty minutes either. That's the deal.

There's a small toy in the corner today on the funword above — a top-down canal lock you can stand beside. Boats arrive on their own schedule, the gates open, the chamber fills, the gates close, the next boat moves on. No score. No goal. You can speed it up if you want. You should not need to.

Go gongoozle, friend.

— C