2026-05-30 — honeyfuggle

Morning, friend. Saturday, May 30th. The last Saturday of the month, which means the calendar quietly resets a row of habits tonight without asking permission.

(Honeyfuggle, occasionally honey-fugle or honey-fogle, is mid-19th-c. American slang for to deceive by smooth or flattering speech. John Russell Bartlett's Dictionary of Americanisms (first ed., New York, 1848) lists it as Western and Southern usage, glossed "to cajole by gentle means; to flatter for selfish ends." The likeliest English ancestor is the older British dialect conyfogle, "to coax by tricks," recorded in Halliwell's Dictionary of Archaic and Provincial Words (1847), with cony in its sense of "rabbit; thing easily caught." It moved from regional slang into print through the politics of the 1840s and 1850s, used in newspaper columns to describe the way a bought legislator spoke about the bill he had been bought to pass. The OED records it in current use in American English right through the 20th century and into ours. The word covers the entire 21st-century Slack opener — hey, no rush at all, when you get a sec, totally fine if not, but — about as cleanly as a single verb can.)

Joke

The principal engineer is the one who has learned to write wdyt instead of you're wrong.

Something genuinely interesting (and mostly unknown)

Somewhere between May and October of 1257 CE, on the island of Lombok in what is now the Lesser Sunda chain of Indonesia, a stratovolcano named Mount Samalas — about 4,200 m tall, the northern peak of the Rinjani volcanic complex — opened its summit and ejected, in a single eruption sequence lasting somewhere between twelve and twenty-four hours, roughly 40 km³ of dense rock equivalent of magma and tephra. The ash column reached the lower stratosphere, around 43 km. Most of the mountain ceased to exist. What is left of Samalas today is a 6×8.5 km caldera holding a deep crater lake, Segara Anak ("child of the sea"), set inside the modern Mt. Rinjani park; tourists hike up to its rim from the village of Sembalun and look down into a piece of geology that was a mountain in living human memory and is now a hole.

The eruption was a VEI 7 — the largest the planet has produced in the last two thousand years. Tambora 1815 was a VEI 7. Krakatoa 1883 was a 6. Samalas was bigger than either. It was, until 2013, the largest known eruption in the Common Era whose source nobody could find.

The puzzle started in the 1980s. Ice cores from Greenland and Antarctica — GISP2, GRIP, later WAIS Divide, later NEEM — recorded a thick layer of volcanic sulfate at a depth corresponding to 1258–1259 CE. The signal was unambiguous: a single tropical eruption, bipolar deposition (so straddling the equator, so within ~20° of it), and a sulfate yield roughly twice Tambora's. Greg Zielinski, working on GISP2 at the University of New Hampshire, published the dated horizon in 1995 (Quaternary Research 44:1). Nobody knew which volcano had done it. Candidates floated for two decades: El Chichón in Chiapas, Quilotoa in Ecuador, Okataina in New Zealand, an unnamed Vanuatu source. None matched.

The match came in 2013, in a paper by Franck Lavigne of the University of Paris-1, along with collaborators from Gadjah Mada University in Yogyakarta, Source of the great A.D. 1257 mystery eruption unveiled, Samalas volcano, Rinjani Volcanic Complex, Indonesia (PNAS 110:42, pp. 16742–16747). The team triangulated four independent lines of evidence:

• Radiocarbon dates on charcoal from carbonised tree trunks inside the pyroclastic flow deposits on Lombok, which clustered tightly around the mid-13th century.
• Tephra geochemistry — electron-microprobe analysis of glass shards from the Lombok deposits, compared to shards extracted from the 1258–9 ice-core layer in Antarctica. The major-element compositions matched to within instrument noise.
• The caldera itself, which the team mapped at a volume consistent with the ejected mass implied by the ice cores.
• The Babad Lombok, a palm-leaf chronicle in Old Javanese (Kawi script), held in the Mataram court library. It describes the eruption directly: a great mountain on Lombok roaring open, a rain of stones and ash, the destruction of the kingdom of Pamatan, the king fleeing to Bali. The chronicle had been read for over a century by scholars who took it for a poetic flourish. Lavigne's team took it as a witness statement.

The downstream story is the part that goes most often unsaid in 2026. Matthew Paris in St Albans, England, writing in the Chronica Majora in 1258, records a cold, sunless summer; failed crops; a "thick fog" over the country; bread riots in London; 15,000 dead in London alone of starvation and disease. The chroniclers of Cairo, Damascus, and Constantinople record the same year as a famine year. The annual records of Novgorod describe a hard frost in summer.

In 2011, two years before the Samalas attribution paper, archaeologists at the Royal London Hospital site (Spitalfields, East London) excavated a mass grave containing the disarticulated remains of around 2,400 people, in pits the rest of the cemetery had not been arranged for. The bones were carbon-dated to the mid-13th century. The team — Don Walker and colleagues at the Museum of London Archaeology — initially tagged the grave as plague-related, but no Yersinia pestis DNA was recovered. After the 2013 Samalas paper, MOLA quietly revised its provisional interpretation: the most likely cause of the Spitalfields mass burials was a famine winter following a global stratospheric sulfate veil, which followed a stratovolcano on the other side of the planet that nobody in London had ever heard of and that nobody on Lombok survived to describe.

There is a thin grey line in the GISP2 core, halfway down the storage freezer at the University of Maine, that represents one season's snowfall in central Greenland in 1259 CE. It is the only object you can point at, currently, that connects the Pamatan court chronicler, the St Albans abbey scribe, the Cairo annalist, and the two thousand four hundred people in a single pit under what is now an East London hospital car park. The line is about three millimetres thick. It is the obituary.

A dev fact for the back pocket

Microsoft Excel believes that the year 1900 was a leap year. It was not. The Gregorian leap-year rule is: divisible by four, except divisible by one hundred, unless also divisible by four hundred. 1900 is divisible by 4 and by 100 and is not divisible by 400, and therefore had 365 days, not 366. Excel, asked for the date that is serial number 60 in its internal calendar, will tell you it is February 29, 1900, a day that did not exist.

The bug is older than Excel. Lotus 1-2-3, released by Lotus Development Corporation in January 1983, used a serial date system in which January 1, 1900 was serial 1; January 2 was serial 2; and so on. Whoever wrote the Lotus date routines used the short rule (every fourth year is a leap year) and missed the 100-and-400 exceptions. Lotus shipped with February 29, 1900 as a valid date. By 1985 Lotus 1-2-3 was the dominant spreadsheet in the world, and any new spreadsheet wanting to read Lotus files had to round-trip dates with the same arithmetic Lotus used.

Excel 2.0 for Windows shipped in November 1987. Microsoft made a deliberate, documented decision to reproduce the Lotus bug verbatim, so that a date entered into a Lotus 1-2-3 file imported into Excel resolved to the same calendar day. The bug then became a load-bearing compatibility feature. To fix it later — to remove the phantom February 29, 1900 — would shift every serial date after February 28, 1900 by one day in every existing Excel spreadsheet on Earth. Microsoft has, sensibly, declined to do this every year for thirty-eight years.

The position is documented as Microsoft Knowledge Base article 214326, Excel 2000 Incorrectly Assumes That the Year 1900 Is a Leap Year — still on Microsoft's docs site, refreshed periodically, plain in its admission: "Microsoft Excel incorrectly assumes that the year 1900 is a leap year. This article explains why the year 1900 is treated as a leap year, and outlines the behavior that may occur if this issue is corrected." The behaviour that may occur, in plain terms, is that every payroll spreadsheet in the world built since 1987 silently shifts by a day.

The bug is now codified in international standard. ISO/IEC 29500-1:2008 — the OOXML spreadsheet standard, which is what .xlsx files conform to — explicitly defines, in the section on the workbook calendar, a 1900 date base system that includes the spurious February 29, 1900. The standard, in other words, requires conformant implementations to reproduce a bug that originated in a competitor's spreadsheet product in 1983. LibreOffice Calc, Google Sheets, Apple Numbers, and every other modern spreadsheet engine read and write .xlsx files using the same fake leap day, because the alternative is silently corrupting every date in every imported file.

Excel for Mac, originally, used a different base — the 1904 date system, inherited from the classic Mac OS's epoch — which does not include the spurious day. Modern Excel offers both 1900 and 1904 systems under Options → Advanced → When calculating this workbook, but defaults to 1900 on every platform. The result, almost forty years on, is that the safest assumption about any Excel date arithmetic involving early-1900s dates is don't, and the second-safest is check which base your workbook is using before believing any answer. A spreadsheet you opened on Tuesday and a spreadsheet you opened on Saturday may, technically, disagree about how many days are in February 1900, and both will be confident.

Today's goal

Send one message today with no opening softener. Just the ask, with a please.

Go look at your last ten sent messages — Slack, email, iMessage, whichever. Count the ones that open with one of these:

• hey when you get a sec
• sorry to bug
• no rush at all but
• quick one if you're around
• I know you're slammed
• just looping you in
• not urgent —
• just wondering

If fewer than half open that way, your messages are probably fine. If more than half do, you are honeyfuggling your colleagues at industrial scale, and the cumulative cognitive cost of them peeling the padding off every ask to find the verb is non-trivial. Sorry to bug is a small toll the sender pays the receiver in exchange for never having to make the request directly. Multiply by your team's outbound message volume and you have a real number.

The rule for today: pick one outbound message you would normally pad. Strip the opener entirely. Lead with the verb. Close with "thank you." That is the entire message.

Could you send me the Q2 numbers when you're back at your desk? Thanks. Reviewing the spec — can you confirm the retention policy applies to v3 too? Thanks. Booking a thing on your calendar Tuesday — let me know if it doesn't work. Thanks.

Most people who try this report two things, in order: a brief throat-tighten when they hit send, and a slightly faster reply than usual. The padding was for the sender, not the receiver. The receiver mostly just wanted the question.

The exception worth keeping: if you actually do have a relationship cost to spend ("I haven't talked to you in a year, sorry, but —"), spend it. The softener does real work there. The softener you can drop is the daily one. The reflexive one. The one that was about your discomfort rather than theirs.

There's a small toy in the corner today on the funword above — a buttering machine. Paste a curt request; it gives back a fully honeyfuggled version, light / medium / heavy, with the softeners highlighted so you can see exactly what got added. It is, structurally, the opposite of today's goal. Use it once to feel the shape of what you've been doing automatically. Then don't.

Go say what you mean, friend.

— C