Clear your diary for Friday, 13 November 2026. According to a paper in Science, that is the day humanity reaches infinity. Nobody expects anything to happen, and that includes the people who wrote it.

The paper ran on 4 November 1960 under the deadpan title “Doomsday: Friday, 13 November, A.D. 2026”. Its lead author, Heinz von Foerster, was a cybernetics professor who had once been a stage magician. With his students Patricia M. Mora and Lawrence W. Amiot, he took 24 estimates of world population covering the past 2,000 years and asked which smooth curve fitted them best. The answer was a hyperbola. Unlike an exponential, a hyperbola does not just grow fast. It reaches infinity at a finite date, which Futura reports as 2026.87, give or take about 5.5 years.

The reasoning behind it was more generous than the title. Humans cooperate and solve problems, so growth could outrun the exponential. The trio then delivered the line that made the paper famous: “Our great-great-grandchildren will not starve. They will be squeezed to death.” Their remedy was plain. Halve the global birth rate.

The world took the bait. Time and The New York Times covered it, and Pogo ran a series of strips. Letters to Science called it “ridiculous” and “manifestly absurd”, and the authors calmly defended their maths. Then the awkward part arrived: the curve kept working. Writing in The Conversation, republished by ScienceDaily on 16 October, Western University historian Alan MacEachern notes that its forecasts beat most demographers’ through the 1970s and 1980s. “Only around 1993 did that accuracy begin to falter.”

What broke the curve was people. Today the global birth rate is slightly less than half the 1960 rate, which is exactly what the paper asked for. The forecast helped make itself wrong. MacEachern calls it “an ordinary scientific article wrapped in an extraordinary title and conclusion”, and adds: “To draw the world’s attention to a serious matter by way of absurdity — that’s a good trick.” Joel Cohen, a mathematical population biologist at Rockefeller University, told Science (as InDaily reports): “I don’t think [von Foerster] was seriously trying to predict the future.” For background, the UN’s 2024 World Population Prospects put us at 8.2 billion, peaking around 10.3 billion in the mid-2080s.

The quiet heroes are easy to miss. Those 24 estimates were the patient work of historians and demographers who reconstructed headcounts for eras without censuses. The magician only drew the line through their labour.

←TODAY: Offices run on fitted curves: span tables, cost-per-m² regressions, energy-use benchmarks, Grasshopper sliders.
→3012: The formulas that survived were the ones that shipped with their own boundaries written on them.
Fulcrum: A curve’s fit quality and its valid range are two different numbers, and only one of them is usually printed.

Here is the lesson on your desk. Every empirical formula is a tool inside its data and a doomsday just past the edge. A cost-per-m² regression built on a decade of Zürich tenders knows nothing about a timber high-rise. A creep curve fitted to five years of tests says little about year forty. A Grasshopper slider whose domain runs to 50 will happily let you drag it to 500. A curve that fits beautifully tells you nothing about where it stops being true. That is a property of curve fitting, and no amount of R² fixes it. It is the first habit drilled in PAZ Academy’s Grasshopper 2 training: a slider without a documented domain is a guess with a nice handle.

Atelier: Most offices inherit spreadsheets whose formulas outlived the people who calibrated them, and AI assistants now generate new fits faster than anyone can check where they came from. The risk is no longer a bad curve. It is a curve nobody remembers the range of. Monday move: set an office rule that no fitted formula, in Excel or in a Grasshopper definition, ships without a cell or panel stating the data range it was fitted on.

Hack: Wrap every fitted formula so it refuses to answer outside the data it learned from. The snippet below runs in plain Python or a Rhino 8 Python component. Store the fitted range next to the function, and let the guard raise an error instead of returning a confident number. Swap in your own span table or cost curve for f.

FIT_RANGE = (2.0, 12.0)  # span in m the table was fitted on
def guarded(f, x, lo=FIT_RANGE[0], hi=FIT_RANGE[1]):
    if not lo <= x <= hi:
        raise ValueError(f'x={x} outside fitted range {lo}-{hi} m')
    return f(x)

From where I sit, decades on, I can confirm that the 14th arrived on schedule. Open your most-used spreadsheet today, find one fitted formula, and write its data range next to it. See you on the 14th.