The eclipse came in from the wrong side

Claim #3 lists four “red flags” and this is the last of them: the failed trajectory alignment during a solar eclipse. The book does not work it out; the argument lives in a video the book links, and it is not the book’s own. It belongs to a named third party with a real instrument, and it deserves to be answered as such.

Whose argument this is

The material runs three deep. The book’s QR reaches an upload on the Shape Debate channel titled sun observations. Most of that upload is not Shape Debate’s: at 24:04 it becomes a rebroadcast of an episode of a show called Hanging On His Words, whose guest is Jeremy McGarry. The argument, the camera, the frames and the slides are all his. The original is What Did the Sun ECLIPSE? | w/ Jeremy McGarry, streamed 19 June 2024 — worth watching in his own words rather than through this page’s summary of them.

So the answer is addressed to Jeremy McGarry, and the credit for the work is his too. It is not the book’s argument, and this page does not treat it as one.

How an eclipse happens

Before any of his measurements, it is worth having the mechanism in front of you, because everything that follows is an argument about which motion does the work.

The chart below was built on this site for a different eclipse — Reykjavík, 12 August 2026 — and built before the event rather than after. Nine days of Moon paths above one horizon, all on screen together. Laid over them are two sets of dots, both taken at the same clock time each day, 17:50 UTC.

Watch what the two sets of dots do. Each set is nine dots, one for each of the nine days, and every dot is where that body stood at the same hour of the clock.

The Sun’s nine days land almost on top of one another, in a clump you could cover with a thumbnail: at a fixed hour it comes back to nearly the same place, sliding down about a quarter of a degree a day as the season turns. The Moon’s nine days march straight across the chart, roughly fourteen degrees of azimuth between one day and the next, running into the Sun’s clump on the 12th and carrying on out the far side at exactly the same pace.

Erie on 8 April 2024 is the same picture from a different spot on the globe, at a different time of year. The arcs sit at different heights, the dots run at a different angle across the frame, the numbers are all their own — but the pattern is the same: a Sun that returns to nearly the same place at a fixed hour, and a Moon that walks past it at its own pace and keeps going.

His camera was pointed at exactly that for weeks, and it recorded it well. The disagreement is entirely about the step after that.

The chart is reproduced here from Where’s the Moon’s Silhouette?, unchanged, where it is set out in full along with what it was used to predict and how that prediction held up. Positions from Astronomy Engine, cross-checked against JPL Horizons; the numbers are baked into the chart rather than computed in the browser, so it is showing the same figures that page was reviewed on.

What he did, which is more than most

He mounted a Brinno TLC2000 time-lapse camera to his house about three weeks before the 8 April 2024 eclipse, left it fixed, and had it take one frame a minute for weeks either side. From that he traced the Sun’s daily path across the frame, and the Moon’s, and established that on eclipse day the two coincided. His slide states the premise plainly:

“Blue line is the same trajectory path of the Sun. The Moon must match exactly at 15:18 on Apr 8 for total solar eclipse. The moon trajectory on March 24, 2024 is approximately the same height as the Sun trajectory on April 8, 2024.”

Nor is the height comparison, as far as it goes: on 24 March the Moon’s arc topped out at 52.4° and the Sun’s on 8 April at 55.4°, so “approximately the same height” is fair to three degrees. But one word in that sentence is carrying more than it can. The Moon does not have a trajectory the way the Sun does. Its arc drops or climbs six or seven degrees a day, and across the fortnight between his two dates it swings through fifty-six degrees, from 74.7° on 18 March down to 18.4° on 1 April and back up to 54.9° on eclipse day:

That swing is not a wander. It is a sweep — bounded and periodic. Across 2024 the Moon’s culminating altitude at Erie stays between 18.3° and 76.3°, and it runs the width of that band and back every tropical month, a little over twenty-seven days. The Sun’s eclipse-day arc, 55.4°, sits inside it. Any level inside a sweep like that is crossed twice a cycle — once going down, once coming back up. The Moon crossed 55.4° twenty-seven separate times in 2024, on average every 13.6 days, and culminated within three degrees of it on 25 days of the year.

Which is exactly what his two dates are. The falling crossing lands between 23 and 24 March; the rising crossing is 8 April itself. Those are the two days inside his twenty-two, and they are not two accidents. They are the two halves of one cycle.

One wrinkle worth stating, since it cuts his way first: how sharp the match looks depends on where in the sweep the level sits. Near the turning points the Moon lingers — it culminates within three degrees of 20° on 64 days of the year — so a low match would smear across a week and impress nobody. At 55° the Moon is in the fastest part of the sweep — a median 5.7° a day and up to 7.7°, against a median of 2.3° down near 20° — so the crossing is sharp and lands on one or two dates. That is why his comparison reads as a near-identity rather than a smudge. It is also, precisely, the Moon migrating at speed.

Nor is his handling of his own equipment. Before arguing, he puts up a frame from the same camera and labels the curved roofline, the curved horizon and the curved road on it “Fish-Eye Lens”, four times over, and sets them aside as artifacts of the lens rather than evidence of anything. That is a person disowning the easiest available argument in his own favour, unprompted, and it should be said out loud rather than passed over.

The step that does not follow

From “the Sun and the Moon share a path that day” he goes to: the body that crossed the Sun should have travelled along that path. It did not travel exactly along it — it came in from below and to one side — and that is what he calls impossible.

He does not leave the reasoning implicit. Forty-six minutes in, he gives the model it rests on, and it is a good, concrete image:

“This cable or wire that these beads move across — that is fixed, it’s not random. If one of the beads is sliding across it, you can’t come at it at a different angle, it has to slide across that cable or the wire. And the blue line across the top is the trajectory of the Sun during the eclipse.”

Spoken, from a machine transcript — see the draft notice above. The premise it states is the one the rest of this page answers, and it is stated more clearly here than the objection to it usually manages.

He is more right than the objection to him usually allows, and the page should start by saying so. Relative to the Sun, the Moon really does move mostly along the shared arc rather than across it. Decomposed at his site on his afternoon, the along-arc part of that motion is between 76% and 98% of the whole, running backwards down the track: the Moon sits ahead of the Sun in the morning and falls back through it. An approach broadly along the trajectory is exactly what happens.

The eclipse does come in near the shared track. The question is only how near, and what “near” is supposed to be.

The number he actually measured

So the disagreement is not about the kind of motion. It is about one angle: how far off the arc the Moon sits as it crosses. Computed for his coordinates, from an ephemeris, with no reference to his footage:

Position angles at the Sun, from straight up, positive toward increasing azimuth. The tilt is measured against the arc as a line rather than a direction, which is why it stays around 35° while the bite itself swings from one side of the disc to the other. Within about ten minutes of totality the two discs are nearly concentric and there is no side to measure, so those minutes are left out. DE421 via Skyfield, topocentric at his printed coordinates. Reproduce with python3 scripts/eclipse_trajectory.py.

Where the fifteen or twenty degrees comes from

That figure has a specific origin, and it is worth following, because the slide it was said over can be measured.

He puts up his time-lapse frame with the trajectory drawn across it, and beside it his Nikon P1000 close-up of the bitten Sun. Then he takes the trajectory line, moves it down onto the photograph, and says it does not match. The co-host supplies the number; he agrees with it and offers none of his own.

Measured off that frame — his curve traced column by column and its tangent taken where the Sun sits, his red line fitted to its own pixels:

His trajectory blow-up is stamped 14:23 and the Nikon frame 14:28; five minutes moves either angle by about a degree. The 7° between his red line and the model’s bite direction is within the roll of a hand-held Nikon, which the photograph does not record.

The second row is where it goes. A fisheye does not render the sky’s angles as the sky’s angles: the arc that runs at 110° on the sky is bent to 97° where his Sun sits, and that is the lens rather than the sky. The Nikon close-up has no such bending. So the two lines he sets against each other are measured in two different projections, and the 45° he gets is not the tilt. The tilt, taken consistently either way, is 36 to 38°. The lens is worth 13° of rotation on its own; about half of that happens to be cancelled again by his red line sitting 7° off the model’s bite direction, which is the sort of thing an unrecorded camera roll does. The two errors are unrelated and it is coincidence that they partly agree.

He is unusually careful about this lens elsewhere. He puts up a frame, labels the curved roofline, the curved horizon and the curved road “Fish-Eye Lens”, and sets all three aside as artifacts — four times over. The distortion he had already identified is the one that follows the line across.

None of which rescues the argument, and the page should not pretend otherwise: correcting for it takes 45° down to 38°, and the premise needs zero. And “fifteen or twenty” is smaller than any of these — it is an eyeball estimate of an angle that measures forty-five on the frame it was said over.

What the argument turns on is the next step: that this tilt should have been zero. That is the premise, and it is the only part that fails.

Why it came from below

There are three motions in this, and the arc is only one of them.

The Earth’s rotation carries the whole sky round once a day at a single rate — Sun, Moon, Orion, all of it together. That is the daily arc. Because it moves everything at once it adds nothing at all to the distance between the Sun and the Moon: it is not that two things on one track cannot overtake, since trains manage it, but that this motion is not the one doing any overtaking. Its contribution to the gap is zero, on any model of what the sky is.

The Sun has a track of its own, about a degree a day. Photograph it from a fixed spot at the same clock time every day for a year and it does not sit still — it draws the figure-of-eight, the analemma. Over the three weeks he was filming it drew eight degrees of that figure, and he measured them.

And the Moon is the fast one, better than twelve degrees a day relative to the Sun. That is the motion that closes the gap. It belongs to the Moon alone and nothing else in the sky shares it.

So the approach runs along the Moon’s own road rather than along the arc, and the only question left is what angle that road makes with the arc, and which side it comes in on.

That has an answer, and it comes from two tilts and a standpoint. The Earth’s axis leans 23.4°, so the Sun’s yearly path — the ecliptic — is not the celestial equator that the daily arc follows. The Moon’s orbit leans a further 5.1° off the ecliptic. These are not the same plane, and neither is the plane of the daily arc, so the approach cannot lie in it. And the observer is not at the centre of any of them.

Measured in the frame where the daily arc simply is the parallel of declination — no horizon, no fisheye, nothing that depends on where anyone stands:

The direction of closing at one instant, not a line fitted through the whole passage — because, as the right-hand column says, from a place on the surface there is no single direction to fit. The middle column is taken at greatest eclipse and the right-hand one at first and last contact. DE421 via Skyfield; rebuild with python3 scripts/build_trajectory_3d_dataset.py.

The Moon’s own orbit adds five and a half degrees more. And the last nine are there because he was standing on the outside of a turning globe rather than at its middle. Being four thousand miles off the centre moves the Moon in his sky by 0.58° at the start of the afternoon and 0.79° by the end — more than the Sun’s own width, either way — and over the passage that displacement swings through 0.59° as he is carried round the axis. The two bodies close on each other by about two degrees in the same three hours. A shift of six-tenths of a degree, at an angle to a closing of two, tips the direction by about nine degrees — and, because the shift keeps swinging, keeps tipping it. That is the arithmetic of the third row, and of its right-hand column.

That last term is a prediction rather than an explanation, because it depends on where the observer stood. Everyone on the 2024 path saw the same eclipse from a different place on the globe, and the model says each of them saw the Moon arrive at a measurably different angle — more from below, the further south:

Millions of people photographed that eclipse. If the angle the Moon came in at is the same everywhere along the path, this account is wrong and his is closer to right. If it slides down the path in the order above, then the amount by which the approach tilts is telling you the size and shape of the thing you are standing on.

And the second row is the easier test, because it needs only one observer. It says the approach direction should turn during the eclipse — by nearly seven degrees at Erie, by less than one at Mazatlán — and anyone with a sequence of frames from a fixed camera has the measurement already. A wire does not turn. Neither does anything watched from the centre of the Earth. Seven degrees is small against the error bars on a hand-measured photograph, so this is a test for a careful sequence rather than a pair of snapshots. His own nine moments are on the edge of it, which is why the page does not claim to have run it.

The measurement is sound and the expectation is not. What he has found is a real thirty-five degrees, tested against a zero that no model of a tilted, spinning Earth ever offered.

And a scale worth noticing

Across the two and a half hours of partial phases, the Sun and Moon together sweep 33° across his fixed frame. Over those same hours the Moon moves 1.1° relative to the Sun. That is a ratio of about thirty to one. He drew a line thirty-three degrees long and asked whether a one-degree motion lay along it.

The approach is tabulated on the slide that says it is missing

One of his slides puts two U.S. Naval Observatory tables side by side — Altitude and Azimuth of the Sun and of the Moon, both for Erie PA at N 42° 08′, W 80° 05′, both for 8 April 2024, sixteen rows each from 08:00 to 18:00. Between them a caption reads:

“The Sun and Moon are on the Same Trajectory”

The tables are sound. Checked row by row against DE421 for the coordinates printed on the slide, every one of the sixty-four figures agrees to 0.08° or better, which is the rounding. Nothing about this data is in dispute.

Computed from the figures printed on his slide, nothing else. The same residual taken from DE421 instead agrees to 0.05° in size and 1–4° in direction at every row.

Read down that last column. The Moon starts the morning three and a half degrees to the right of the Sun — seven Sun-widths — closes on it all day, arrives at 15:18, and then opens out again on the opposite side. By six in the evening it is a degree and a third above where it began three and a half degrees beside.

That is the approach. It is in his own numbers, on his own slide, under a caption saying the two are on the same path.

Both things he says are true at once, and that is the whole confusion in one picture. The two bodies are on the same trajectory — the daily arcs match to a few tenths of a degree all day, which is what his caption is reading. And the Moon is crossing the Sun — the residual between those matching arcs runs from three and a half degrees on one side to a degree and a third on the other. The trajectory is what they share; the residual is what one does relative to the other. Only the second can cause an eclipse, and it is the column his slide does not compute.

It also settles the question the photographs were raised to decide, without needing them. His table has the Moon to the lower right of the Sun through the afternoon and above it afterwards. That is the bite on the lower right before totality and at the top after — the same rotation the section below predicts from an ephemeris, arrived at here from nothing but his own two columns and a subtraction.

He built the right instrument, and stopped one layer short

This is the part worth being careful about, because the instrument is not the problem. A chart of exactly this kind is on this site already, built for a different eclipse before the event: on Where’s the Moon’s Silhouette? there is a plot of nine daily Moon paths above Reykjavík’s horizon, 8 to 16 August 2026, made to predict the total eclipse of the 12th. Same idea as his: fix the observing point, draw the daily arcs, look at where the two bodies fall.

The difference is one layer of marks. That chart carries a second set of dots — where each body stood at a fixed clock time, 17:50 UTC, one dot per day. And those dots are where the answer lives:

The Sun’s nine dots span eight-tenths of a degree — they pile up on one another, which is what “the Sun returns to about the same place at the same hour” looks like when you draw it. The Moon’s nine march right across the chart, run into the Sun on the 12th, and come out the other side at the same steady pace. It closes on the Sun at about 13½° a day — 55° out, through zero, and 52° out the far side. That closing is the approach, sampled once a day instead of once a minute. The distance it covers across the chart is a larger number, 112° of azimuth, but azimuth is a chart coordinate rather than a physical angle and it comes out differently at every latitude. The closing rate does not: at Erie the same march runs 131° of azimuth and still closes at 13.8° a day.

Measured rather than eyeballed, on that chart: the daily arc at totality runs at 116° and the day-to-day dot track at −88°. Those are 156° apart, which is close to opposite — the dots run back down the arcs rather than across them, with a tilt of a couple of dozen degrees off the line. At Erie the same pair of directions is 143° apart, and the tilt of the bite off the arc line comes out at about thirty-five. The two sites do not give the same number because the projection into a local horizon depends on where you stand and how high the pair sits. What they share is that the approach is never in the arc. Sampling makes no difference either: the Reykjavík approach measured over ten minutes is −93° against −88° measured over two days.

He plotted the arcs and found them the same height, which they were. What he did not plot is the Moon walking along that shared arc toward the Sun, a dozen degrees a day.

He had three weeks of frames at one a minute. The fixed-hour marks were in his own data the whole time, and they are the measurement his argument turns on. We are not saying he would have drawn our conclusion from them — he might read the same dots a different way, and this page cannot speak for that. We are saying the quantity was his to plot, and the instrument he built was good enough to show it.

What the model says his own photographs should show

This is the part that can be checked rather than argued, because he photographed the Sun at nine moments through the event and has the frames. The direction of the bite is set by where the Moon’s centre sits relative to the Sun’s — the notch appears on the side the Moon is coming from — and that direction is not a constant. It rotates.

His Nikon P1000 frame at 14:28 has the bite at the lower right. So does the line above it. The observation he presents as impossible is the one the model puts there — we did not adjust anything to make that come out, and the computation makes no reference to his photograph.

The bite direction turns through about a hundred and sixty degrees across the event. No straight line can do that. It is the wrong kind of object to hold an eclipse against.

And that gives the page a prediction rather than a rebuttal. He took four frames after totality. In every one of them the bite must be at the top of the disc, on the opposite side from the four before. If his own set shows that, the rotation is in his data and was in it all along. If it does not, this page is wrong and we would want to know.

His own footage records the eclipse, to the minute

He published two stretches of unedited camera footage inside the broadcast. The second one runs from midday on 7 April to the evening of the 8th on his camera’s own burnt-in clock — which means the eclipse is inside it.

It is there, and it is not subtle. Through the whole of that afternoon the sky carries thousands of saturated pixels. In one window, and nowhere else in daylight, the count falls to zero:

The sky says when that happened. Computed from DE421 for his printed coordinates, totality at Erie runs 15:16:23 to 15:20:06 EDT, greatest at 15:18:15, lasting 223 seconds. Requiring his dark window to lie inside that brackets his camera’s clock error at 40.6 to 42.2 minutes.

This is the one measurement on the page that fixes an absolute time without borrowing anything from him. Everything else — his boresight, his trajectory, the plate solution below — had to take his stated clock on trust, because a camera pointed a little further east and a clock running a little slow do the same thing to a photograph and no amount of sky can tell them apart. The eclipse can. It happened when it happened.

Solved on the Sun, his camera finds the Moon

The first stretch of raw footage runs from 23 to 25 March. Every distinct frame in it was searched for bright compact sources, and those sources sorted into tracks by nothing but how they move — a cloud edge sparkles for a frame, a streetlight sits at one pixel for three days, and the Sun and the Moon do the one thing neither of those does, which is travel a long smooth path a few pixels a frame for hours on end. Six tracks came out, and only six. No ephemeris was consulted while choosing them, so the measurements cannot have been picked for agreeing with the model.

That is 1,256 positions: 722 of the Sun and 534 of the Moon, across the full width of his frame. His camera was then solved on the Sun alone — where it points, how long its lens is, how much its fisheye bends straight lines. That solution never saw the Moon.

The Sun row is 556 of the 722, the other 166 being frames where cloud outweighed the disc or the disc sat in a corner the lens model does not reach; they are rejected rather than silently dropped, and the count is published here for the same reason. Nothing is rejected from the Moon. The lens order was chosen by the body the fit never saw rather than the one it did: one distortion term beats two, three and four on the held-out Moon. The 7 April row uses the March solution unchanged, with the clock taken from the eclipse above; between the two dates his own landmarks confirm the camera had not moved — the isolated spruce tip and the left roofline both align to within three pixels.

The same model, walked forward a day at a time

Everything above is one comparison at a time: his camera against the ephemeris on three nights, and the eclipse on one afternoon. It is worth seeing them as one motion, because the argument is not about any single day. It is about what the Moon does between days.

Below, the model draws the Sun’s arc and the Moon’s over Erie for each of the twenty-two days from 18 March to 8 April, and runs a clock through each day with both bodies moving along their arcs at the same instant. That shared clock is the point of the thing. A chart of altitude against azimuth with no time in it will happily draw the two arcs on top of each other whenever they cross the south at a similar height — which on 25 March they do, twelve hours apart — and a reader looking at that picture would see the very coincidence this page is taking apart.

Each day leaves a dot at the Moon’s highest point, so the fortnight’s swing accumulates as a column rather than a claim. The readout counts down how far the Moon still has to travel to reach the Sun: 250° on 18 March, falling by ten to fourteen degrees a day, 0° on 8 April. On the four days his footage can be measured — 23, 24 and 25 March, and 7 April — his own positions appear along the arcs as the clock reaches them.

Eclipse afternoon itself is not on the chart, and it should be said why — because the reason is not that his camera failed. Before totality Erie is under broken cloud: twenty to forty-six bright patches a frame with the Sun mostly behind them, and no smooth track survives. After totality it does. From 15:45 the same solution follows the Sun back across the sky for an hour and a half, and puts it a median 1.5° from where the ephemeris says it was, with 86% inside two degrees. That is a real recovery of the Sun on the disputed afternoon, and it agrees with the model.

It is still not to the standard the rest of this page holds, and the reason is specific. The residual is systematic, not scatter: the measured Sun sits consistently inside the predicted one, by about 5 px at mid-frame and 16 px out near the corner. That is the single radial term in the lens model under-bending at large radius — the same corner the March fit already rejects points from. By 17:19 the Sun is predicted outside a 1280-pixel frame altogether and what the tracker follows after that is cloud. A degree and a half of lens error is not a measurement to put beside 0.63°, so the positions stay off the chart while the numbers stay in the repository. Rebuild with python3 scripts/apr_track2.py and scripts/apr_export2.py; positions at data/erie-eclipse-day-pixels.csv.

The same afternoon, from outside

Everything so far has been drawn on the sky — arcs, angles, a direction of approach. That is the right frame for answering him, because it is the frame he measured in. But it leaves out the thing that was physically happening, and the two are worth seeing together.

Below, the same event drawn twice. From the side, a cross-section: the Sun far off to the left beyond the Moon, the Moon’s shadow narrowing to a thin cone, and the patch of it that lands on the sunlit face of the Earth. From outside, the same thing in three dimensions on a turning globe, with the path of totality drawn on the surface. One clock runs both. Watch it through and the shadow arrives, crosses, stretches out as it goes and finally slides off the Earth altogether — while Erie is carried round into it and out again by the Earth’s own rotation.

The instrument is sound, the pointing is sound, the clock is sound, and the model that says where the Moon was on those nights is the same model that says where the bite sat on 8 April. There is no version of this where the framework is reliable in March and unreliable on the one afternoon its answer is unwelcome.

What would change our mind

• His post-totality frames showing the bite still on the lower right rather than the top. That would break the account above directly, and it is his footage that settles it.
• A worked derivation in which the approach direction of an eclipsing body is predicted from the shared diurnal arc. We do not think one exists, but it is what the claim needs.
• A totality time materially different from 15:18 at his coordinates, which would mean the ephemeris used here does not describe his sky.
• Photographs from elsewhere on the 2024 path showing the Moon arriving at the same angle it did at Erie. The account above says the angle slid from about 45° in Mazatlán to 35° in Montreal, because the observer’s own place on the globe is a term in it. Millions of people photographed that eclipse; if the angle is flat along the path, that term is not real and this page has overclaimed.

Where this page could be wrong

• Parallax is not in the account above, and it is not negligible. The Moon’s position seen from a point on the Earth’s surface differs from its position seen from the centre by up to about a degree — more than the Sun’s own diameter, and the reason eclipse tracks are narrow strips rather than bands. The figures in the tables are topocentric and so include it, but the argument above is told without it, and a reader who knows eclipses will notice.
• His camera’s roll is assumed. “Lower right” is a claim about a level frame. If the Nikon was rotated when the 14:28 frame was taken, the match to his photograph is weaker than stated — though the rotation of the bite through the event does not depend on it.
• The strongest form of his argument is reconstructed from a transcript and sixteen frames, not from watching the hour. If he makes it more carefully somewhere in the stretch not examined closely, this page is answering a weaker version than the one he offered.
• “15, 20 degrees” is a co-host’s estimate off a slide, not a figure Jeremy McGarry claims. Nothing here rests on it and the page does not use it as his number. It is also the page’s one spoken quotation and comes from a machine transcript.
• The dark window is read as totality because of what the frames show and when they show it — the Sun reduced to a point, twilight colour on both horizons at once, and the whole thing over inside four minutes. A thick enough cloud crossing the Sun at exactly the computed minute would look different in the horizon glow but similar in the pixel counts, and the pixel counts are what the table gives. The frames are reproduced above so a reader can disagree.
• The angles measured off his slide are measured off a screen recording of it, at 1280 × 720, three generations from the original. A degree or two either way is not recoverable. Nothing in the section turns on a degree or two.
• The nine degrees the page attributes to standing on the surface is the newest claim here and the least cross-checked. It is a robust output of the ephemeris — the geocentric track is steady to a tenth of a degree while the topocentric one moves with the observer — but it has not been checked against anyone’s actual photographs from a second site, which is exactly the check it invites.
• The plate solution puts the lens’s axis 23 pixels right of the frame’s centre. That offset is what reconciles the fitted boresight with his published 187.9°, and it is a fitted quantity rather than a measured one. If it is wrong, the frame-centre figures above stand and the fitted ones move.

How to check this yourself

Everything in the two tables comes out of one script with no reference to any of his material:

python3 scripts/eclipse_trajectory.py

It takes a latitude, a longitude and a date, and prints the two motions, the sweep-versus-approach ratio, and the bite direction at each moment. The only inputs are his published coordinates and the ephemeris. There is no fitted parameter anywhere in it, and nothing in it was tuned to reproduce his photograph — that it does is the result, not the method.

Sources & further reading

• The argument is Jeremy McGarry’s, presented as a guest on What Did the Sun ECLIPSE? | w/ Jeremy McGarry, streamed 19 June 2024 on Hanging On His Words. The copy actually consulted here is a rebroadcast within an unlisted upload titled sun observations on the Shape Debate channel, which the book’s own QR reaches; the embedded portion begins at 24:04. The two recordings have been compared and are the same broadcast, the re-upload slightly trimmed. Quotations on this page are read off the slides in frame rather than taken from a transcript, with one exception marked as such in the draft notice above. Figures and measurements attributed to his footage are taken from the two raw camera segments played within the broadcast.
• Positions computed from JPL’s DE421 ephemeris via Skyfield, topocentric for 42.129° N, 80.085° W.
• Levi Miller, Globe Deconstruction? (prerelease) — p. 166 (the four red flags, of which this is the fourth), p. 167 (the QR codes).
• Cross-references on this site: Where’s the Moon’s Silhouette?, Full-Moon Lighting, The Moon-Tilt Illusion and Jupiter’s Shadows — the other limbs of the same claim. The first half of the same video argues that sunset is perspective convergence rather than occultation; that is answered at Perspective Cannot Make a Sunset and The Sun Does Not Shrink.