A Mark on the Glass.

For nineteen releases we ran honest math on other people’s mysteries. For the twentieth we did the thing the series is actually for: we put a real plate on the bench. On a red photographic plate exposed at Palomar on 12 April 1950 — plate XE 325 — nine star-like points sit in a tight 10′ patch of sky. They are not on the blue plate taken thirty minutes earlier, not on the red plate six days later, and not in any survey since. We fetched the digitised plate from the archive and measured the dots ourselves.

Where we land: still open — and earned by the pixels, not by opinion. The marks are real (we watched them vanish in data we fetched ourselves). Their shape is round and star-sized (we measured it) — and roundness settles nothing, because the plate kept the shape and threw away the clock. The one clue that reaches past the glass is itself contested. No god number — a god hold.

1 · The plate is the medium

Start with the frame, because it decides everything. Our reading of the world is dark = the consistent (what is truly there) seen through light = the medium of observation. A photographic plate is that medium frozen on glass — captured light, chemically fixed. So the first question about a bright dot is never “what is it in the sky?” It is the question Release 001 asked of the Apollo “UFO” that turned out to be a cosmic-ray track struck into the film itself: is the mark in the sky, or in the film? A mark on the glass is not yet a fact about the sky. Check what you brought.

The VASCO project (Vanishing & Appearing Sources over a Century of Observations, led by Beatriz Villarroel) found these dots by comparing the 1950s Palomar plates to modern surveys. We reproduced the core observation on data we pulled ourselves: the real POSS-I red cutout of the field, and the POSS-II red cutout of the same sky decades later.

2 · We measured the dots

The whole public fight is about one thing: the dots look round, like real stars. So we measured it. On the real scan we ran our own shape fitter over the transient sources and a control group of forty field stars matched in brightness on the same plate — after first checking the instrument could recover an ordinary star at the paper’s own image scale, because a tool that has never graded a known case has no business grading an unknown one. The two numbers that matter: FWHM — the width of a star’s blur — and ellipticity, how squashed that blur is.

That one measurement reproduces both camps at once — which is the point. The skeptics’ anchor, Hambly & Blair 2024, finds all nine sit on the concentrated, circular tail of the profile distribution, and machine-classifies all nine as “bad” detections at 62–99% confidence: rounder than a star, so an emulsion flaw. The pro-transient rebuttal — Villarroel, Solano & Marcy 2025 — answers that a sub-second flash is naturally rounder and sharper than a star: a 40–50 minute tracked exposure smears every real star by 1–3″ of atmospheric shimmer, shake and tracking drift that a frozen instant escapes. Both are correct. Round is compatible with a defect and a flash and an ordinary source. The most argued-about feature in the whole case cannot tell the sides apart.

And here is the honest caution on our own measurement: we measured a digital scan of a copy plate — the same class of data Hambly & Blair used, and the same class everyone agrees cannot settle it. Across two decades of optical-flash searches on ~910,000 archival plates, not one candidate flash was ever confirmed as real rather than a defect — because the deciding tests (halation rings, reciprocity failure, saturation under a microscope — the chemistry-level fingerprints) live on the physical glass, which VASCO does not have and neither do we. Villarroel says it plainly: “we have no access to the original POSS-I plates.”

3 · The plate threw away the clock

Why can’t the shape decide? Our framework has a one-line answer, and here it is not a metaphor — it is literally what the instrument did. A long exposure keeps the event-shape and discards the clock. Fifty minutes of collecting light integrates away the time inside it: the plate records that a dot was there, and destroys the information of how long it was lit. A source that flared for half a second and one that glowed the whole exposure land as the same round grain, differing only in brightness.

The plate kept the shape and threw away the clock.

That is the whole roundness debate in one line: it can only exist because the medium discarded the clock. And roundness is symmetric evidence — it knocks out “round means defect” and “round means real” in the same stroke. So we turn: if the shape can’t decide, what can?

4 · The line — one object, or nine?

Six of the nine points fall on two straight lines of three. Villarroel notes this “as a curiosity” and models the field as nine independent glints. But the more economical reading has never been knocked down: one tumbling, reflective object crossing the field and flashing repeatedly draws a collinear string of round points for free — one object, not nine, with the line as its signature rather than a coincidence. Occam prefers the one over the nine.

We’re not calling it — and we won’t let it be counted twice. The reported alignment significance (2.5–3.9σ, where σ is astronomy’s odds-of-being-a-fluke unit) is computed against a null of many independent random points, so it cannot tell one object from nine — it is not extra evidence of a crowd. And the one-object reading carries two real problems, stated: nothing artificial is known to have orbited before Sputnik (1957, seven years after the plate), and the faint modern galaxies sitting at these positions fight a fast-moving object. An open guess, flagged as a guess — the economical reading, not the proven one.

5 · The clues that couple to the sky

Two features of the wider ~100,000-transient catalogue do reach past the glass — a flaw in an emulsion cannot know either of them — and they are the strongest cards on the table. They also point in different directions, which is exactly why the case stays open.

The Earth-shadow deficit is the exotic card: fewer transients fall inside Earth’s shadow cone than chance allows — the signature of things that need sunlight, meaning things at altitude, not marks on glass. It shows up in the VASCO team’s model and in an independent reanalysis of their catalogue (Doherty 2026, a preprint) — but it is the exotic case’s load-bearing beam, and Watters et al. 2026 knock it out: under a corrected, non-uniform background the deficit is not statistically significant. The strongest evidence for objects-in-orbit rests on a disputed baseline.

The nuclear-test correlation is the famous one: transients are ~45% more likely within a day of an atmospheric bomb test, peaking the day after (Bruehl & Villarroel 2025, peer-reviewed). But read the mechanism, not the headline. The most natural cause of “more false stars right after a bomb” is radioactive fallout dusting the plates — an effect in the film, not the sky — and the roughly one-day Nevada-to-Palomar fallout journey fits the day-after peak and has never been quantitatively ruled out. The correlation also never controlled for which nights Palomar was actually observing. The spookiest-sounding clue points, mechanically, straight back at the glass.

The counter-weight — the mundane case at full strength

We say the hard parts out loud, because that is the deal. The skeptic case is strong. Watters et al. 2026 show the working catalogues were never cleaned of scan artefacts (up to ~40% suspect; ≥91% of one set within 5″ of a catalogue star), and estimate the plate set could hold hundreds of star-like emulsion faults per plate by chance alone. arXiv declined the two October-2025 VASCO papers as lacking “substantive scholarly research”; the lead author on the nuclear statistics is a pain psychologist, not an astronomer; and every professional astronomer quoted in mainstream coverage favours a mundane reading — coincidence, atmospheric disturbance, or plate artefacts. The decisive fact remains that the case-closing tests need the physical glass, and no one has it. This is not a thin mystery bravely defended; it is a genuinely unresolved anomaly with the smart money on the mundane.

Why our math sees more

Because we fetched the plate instead of narrating it — and because the framework was the test, not the decoration. Before the pixels arrived, it told us where a photographic anomaly goes to die: the medium keeps the shape and discards the clock, so shape can never single out a cause on a plate that integrated for fifty minutes. We measured, and it couldn’t. When the answer is a non-answer, the honest move is to measure it precisely and hand you the wall, not a story painted over it.