--- slug: eclipse-here passed: 2026-11-16 by: "a claims-pass subagent of claude-busy-maxwell-lgas9j, which wrote no part of the page, the entry or the source record; one agent on oversight/claims-pass.md, before publishing; all 33 entries of assay/sources/eclipse-here.json and 40 claims it leaves out, at 510e66e973 on 15 November 2026. Brought up to date at c1a89dfe80 on 16 November 2026 by a second such subagent, which also wrote none of it: all 35 entries and 50 claims they leave out. Brought up to date at dff219b53b on 16 November 2026 by a third such subagent, which also wrote none of it: every line read again against that commit, and 13 claims added for what it says that is new or that the record had left out. Re-checked at a49593092d on 16 November 2026 by a fourth such subagent, which also wrote none of it: the five findings the third left open (17, 21, 22, 23 and 24) and the five lines they belong to, each checked afresh, the published check run from an empty folder as a reader would, and one more corner found (finding 26). Re-checked at c3ad3a26e8 on 16 November 2026 by a fifth such subagent, which also wrote none of it: finding 26 and its line (card-refuses), with the page loaded in a browser at four addresses under its script as committed and again under a49593092d's, and the one other line that cites the changed code (limb-map) brought into step" checked: 98 confirmed: 94 wrong: 0 unverifiable: 0 observed: 4 stale: 0 --- Read before publishing by five readers who wrote none of the page, the entry or the source record: first on 15 November 2026, while the author was still correcting the page on other reviewers' reports; again on 16 November at c1a89dfe80, after the author had changed the page, the entry, the source record and the engine in answer to the first pass and to four other reviews; a third time at dff219b53b (16 November, 01:41 UTC), after the author had taken this record's four open items and a house-rules review into the page, the entry, the Tools card, the source record, the engine, the check and the narration; a fourth time at a49593092d (16 November, 02:21 UTC), after the author had taken the five items the third reading left open into the check, the page's script, the entry, the source record and the Tools card; and a fifth time at c3ad3a26e8 (16 November, 02:43 UTC), after the author had taken the one item the fourth reading left open into the page's script. The lines below describe the page, the entry, the Tools card, the source record, llms.txt and the TV narration as committed at c3ad3a26e8. On those surfaces that commit changes one line of the page's script, the test that shows the note pointing to real-limb maps (finding 26), and puts a comment above it; so for this update the line that finding belongs to (card-refuses) was checked afresh in a browser, and the one other line that cites that test (limb-map) was brought into step. The other 96 stand as the fourth reading left them at a49593092d, which on those surfaces changed only the sentences of findings 17, 22, 23 and 24 and the two identical copies of the check (finding 21): that reading checked afresh the five lines those findings belong to (check-keys, run-anywhere, sunset, limb-map and card-refuses) and brought one more (keys-and-browser) into step with the entry, and the rest stand as the third reading left them, at dff219b53b, when every line was read again, the check's keyed figures were compared with c1a89dfe80's (all 440 the same), and 13 lines were added: for the statements new in dff219b53b, for the narration, which the record had not covered, and for the page's promise that its check runs from an empty folder. A claim that was false or loose when first read is recorded as it now stands, and its finding says what it said then: 5 were WRONG and 11 MINOR at the first pass (findings 1 to 16), all corrected; the 4 MINOR left open at c1a89dfe80 (17 to 20) are corrected, three at dff219b53b and the last at a49593092d; five things this record had passed or left out at c1a89dfe80, which the house-rules review found, are corrected (finding 25); the WRONG one and the three MINOR ones new at dff219b53b (findings 21 to 24) are corrected at a49593092d; and the MINOR one found at a49593092d, in a sentence of the Tools card, is corrected at c3ad3a26e8 (finding 26: where the reader's own switches were far from every named set's, the page could refuse a verdict without the pointer to a real-limb map that the card promises for 2 August 2027 and 22 July 2028). 98 lines: 94 confirmed and 4 first-hand observations. No finding is left open. How it was checked at c1a89dfe80. The repository's check, node verify-eclipse-here.mjs: exit 0, 122 passed, 0 failed, with a browser. On a scratch copy holding only the served files: 118 passed, 0 failed; with one keyed figure changed in a sentence it fails (5 of 5 tried), and with an unkeyed one it passes (7 tried). The three Python programs, run again outside the repository on JPL's de440s.bsp, give elements-de440.json, check/de440-check.json and check/topocentric-check.json byte for byte. The second ways: the first pass's Besselian calculator (written from Meeus and the Explanatory Supplement, not from the page's engine) and its direct computation on DE440 with the light time taken to each observer, run again; this pass's own computation of the central line from Meeus's formulas; and Skyfield on DE440 for Moscow's sunset, for every solar eclipse from 2027 to 2035 and for the direction of the drawn Moon. The page was loaded in Chromium from a local copy, with every request to anyone else blocked, at fourteen addresses. Not done: the office's online look-up of each excerpt (node assay/sources.mjs eclipse-here --online), because this pass was told to keep off the network; the offline checks report the record sound, and each excerpt was found in its kept copy. Not yet published, so the reader's download was not tried. How it was checked at dff219b53b. git diff c1a89dfe80 dff219b53b over the page and its modules, the entry, the source record, the card, llms.txt, the narration and the check, read line by line. The repository's check, node verify-eclipse-here.mjs: exit 0, 122 passed, 0 failed, with a browser, in 85 s; node verify-eclipse-here.mjs --facts prints the same 440 figures as at c1a89dfe80, every one. On a scratch copy holding only the served files: 118 passed, 0 failed; a keyed figure changed in a sentence fails it (2 of 2, one of them in the new sentence on the elements), an unkeyed one does not (2 tried), and of the cells the page's script fills in, the when and edge cells of the table of named sets and the times in UT fail it when altered (3 of 3) while the Sun's altitude and the second clock do not (2 of 2). Copied alone into an empty folder, as the page tells a reader, the published check stops at once (finding 21); with two of its lines moved, it downloads and passes. The second ways: a Besselian calculator written for this update from Meeus, sharing no code with the engine or with the earlier passes' calculators, for Moscow's sunset, a sunrise and the two radius switches at NASA's points of greatest eclipse; Skyfield on DE440, with the Sun and Moon as the observer sees them, for Moscow; the engine itself at greatest eclipse of all 16 central eclipses; the third computation's crossings recounted by the Sun's altitude; and the kept copies, for the two pages now in the list of sources and for the page's four quotations, each found again character for character. The page was loaded in Chromium from a local copy, with every request to anyone else blocked, at 16 addresses, and at the bare address once more with a device position given to it. Not done: the office's online look-up of each excerpt (node assay/sources.mjs eclipse-here --online), this pass having been told to keep off the network; the offline checks report the source record sound. Not yet published, so the reader's download was not tried against the site. How it was checked at a49593092d. git diff dff219b53b a49593092d over the check, the page and its modules, the entry, the source record and the Tools card, read line by line. The reader's run: public/checks/verify-eclipse-here.mjs, identical to verify-eclipse-here.mjs, copied alone into an empty folder and run with Node 22 as the page says, with two modules loaded first sending https://artwaste.land to a local server over public/, logging each request and refusing any other: 14 downloads, 118 passed, 0 failed, exit 0, in 39 s, and no /tmp/eclipse-here-* folder left; with every download answered 404 it stops at the first, exit 1, and leaves none either. The repository's check, node verify-eclipse-here.mjs, inside the repository: exit 0, 122 passed, 0 failed, with a browser, in 83 s, and the working tree unchanged. The page loaded in Chromium from a local copy, with the place look-up blocked (logged at three of the addresses, it asked nothing of anyone but the local server), at nine addresses: Moscow on 12 August 2026; 36° S, 153° W on 8 April 2024; Stephenville; just inside the northern edges of 6 February 2027, 26 January 2028 and 22 July 2028; the bare address; Sydney's meridian on the centre line of 22 July 2028; and 32.37° S, 151.2093° E on 22 July 2028 with a Delta T of 200 s. The second way for the two horizon sentences: Skyfield on DE440, the Sun and Moon as the observer sees them, NASA's Delta T and no refraction. The entry's and the source record's sentence read against what app.mjs marks and family M compares. The change proposed in finding 26 tried on a scratch copy of the served files, the check run on it. Not done: the office's online look-up of each excerpt, this pass having been told to keep off the network (no excerpt changed at a49593092d, and the offline checks report the source record sound). Not yet published, so the reader's download was not tried against the site. How it was checked at c3ad3a26e8. git show --stat c3ad3a26e8: the page's script and this record, nothing else; git diff a49593092d c3ad3a26e8 over public/: public/tools/eclipse-here/app.mjs alone, its line 122 replaced by the line finding 26 proposed (now line 123) under a comment saying when the note is shown, the code the same as on the copy the fourth reading tried. The page loaded in Chromium from a local copy of public/, every request to anyone but the local server refused and logged (none was made, and there was no script error), at four addresses: 32.37° S on Sydney's meridian on 22 July 2028 with a Delta T of 200 s, the address of finding 26; the bare address; Sydney's centre line on 22 July 2028; and, chosen for this reading, 36.59° N, 4.8431° W on 2 August 2027 with a lunar radius of 0.27 and a solar radius of 965″, where the reader's own switches put the place 1.55 km outside the edge and each named set puts it 21.67 to 23.96 km inside. At each were read the verdict, every named set's distance to the edge, whether the note under the verdict (#limb) was hidden, both as the attribute and as drawn, and what its pointer (#limb-map) said and linked; then the same four with a49593092d's app.mjs served in place of the committed one. At the first address the switches were then changed on the page as a reader would, to NASA's named set and back to a Delta T of 200 s, and the note hid and showed again with them. Not done: the page's check, not run for this update, since it names neither the note nor its pointer and neither copy has changed since a49593092d (the fourth reading ran it on a copy of the served files holding the same code without the comment: 118 of 118); the office's online look-up of each excerpt, this pass having been told to keep off the network (no excerpt changed at c3ad3a26e8, and the offline checks report the source record sound). Not yet published, so the reader's download was not tried against the site. ## Claims - CONFIRMED | [four-things] Every calculator that treats the Moon as a smooth ball has to fix four things before it can answer for a place: which elements, what Delta T, what radius for the Moon and what radius for the Sun. | public/tools/eclipse-here/eclipse.mjs | The engine's atPlace() takes exactly these: a set of elements, and deltaT, k1 and k2, sunSD, and now refuses to run without all four. The first pass's calculator needed the same four, the place and its height, and nothing else to give the page's figures back. The entry now also names a fifth convention, the centre-of-figure correction, which NASA chooses not to apply: an option, so "has to fix four" still holds, and the page's lede says "at least four". - CONFIRMED | [nasa-ephemerides] NASA's page of elements for the eclipse of 2 August 2027 says its predictions were generated using the VSOP87/ELP2000-85 ephemerides and a Delta T of 71.7 seconds. | https://eclipse.gsfc.nasa.gov/SEbeselm/SEbeselm2001/SE2027Aug02Tbeselm.html | Kept copy: "Predictions for the Total Solar Eclipse of 2027 Aug 02 were generated using the VSOP87/ELP2000-85 solar and lunar ephemerides and a value of ΔT = 71.7 seconds." All 22 kept pages name the same ephemerides. - CONFIRMED | [worth-elements] On DE440 the path of 2027 lies 0.27 km south of NASA's in southern Spain and the path of 22 July 2028 lies 0.67 km south at Sydney's meridian, each with its width unchanged. | verify-eclipse-here.mjs | Family H at c1a89dfe80 and again at dff219b53b: 0.27 and 0.67 km south, widths 0.00 and -0.01 km. The first pass's calculator, run again on the two files: the edges move -0.27 and +0.27 km in Spain, -0.67 and +0.67 km at Sydney's meridian, widths +0.00 and -0.01 km. WRONG as first read (0.23 and 0.56 km, from elements without the Sun's aberration): findings 1 and 2. - CONFIRMED | [iers-prediction] IERS Bulletin A of 12 November 2026 predicts UT1 minus UTC of -0.24398 s for 2 August 2027, which puts Delta T at 69.428 s. | https://maia.usno.navy.mil/ser7/ser7.dat | Kept copy, headed "12 November 2026 Vol. XXXIX No. 046": the row "2027 8 2 61619 0.2635 0.3816 -0.24398", with "TT = TAI + 32.184 seconds" and "TAI-UTC = 37.000 000 seconds". 32.184 + 37 + 0.24398 = 69.42798. - CONFIRMED | [dt-all-larger] For every one of the 20 eclipses still to come, NASA's page assumed a larger Delta T than the Bulletin now gives. | verify-eclipse-here.mjs | Recomputed here without the engine from the kept copies: NASA's 22 values read off its pages, the Bulletin's daily table for the two eclipses it reaches and its formula for the other 18. NASA's is larger in all 20, by 2.3 s (2027) to 9.8 s (September 2035). - CONFIRMED | [iers-extension] Past the end of its table the Bulletin offers a formula which it says "will not reproduce the predictions given below, but may be used to extend the predictions beyond the end of this table". | https://maia.usno.navy.mil/ser7/ser7.dat | Kept copy, under PREDICTIONS: "The following formulas will not reproduce the predictions given below, but may be used to extend the predictions beyond the end of this table." Character for character across its line break, in the entry and on the page. - CONFIRMED | [worth-deltat] Changing only Delta T moves the edge of the 2027 path by 0.03 km in southern Spain, where the path runs east to west, and the edge of the 2028 path by 0.65 km at Sydney's meridian; in both places the clock time of maximum moves by about three seconds. | verify-eclipse-here.mjs | Family H at c1a89dfe80 and again at dff219b53b, and the first pass's calculator run again: northern edge +0.03 km in Spain and -0.65 km at Sydney's meridian; maximum 3.0 s and 3.4 s later. - CONFIRMED | [nasa-two-radii] The same NASA page gives two lunar radius constants, k1 = 0.272488 for the penumbra and k2 = 0.272281 for the umbra. | https://eclipse.gsfc.nasa.gov/SEbeselm/SEbeselm2001/SE2027Aug02Tbeselm.html | Kept copy: "Lunar Radius k1 = 0.272488 (Penumbra)" and "Constants: k2 = 0.272281 (Umbra)"; the same on all 22 kept pages. - CONFIRMED | [iau-radius] In August 1982 the International Astronomical Union adopted k = 0.2725076 for the mean lunar radius. | https://eclipse.gsfc.nasa.gov/SEmono/reference/radius.html | Kept copy: "In August 1982, the International Astronomical Union (IAU) General Assembly adopted a value of k = 0.2725076 for the mean lunar radius." - CONFIRMED | [worth-moon] Changing only the Moon's radius to the IAU's value moves each edge of the 2027 path outward by about a kilometre and a half and lengthens totality on the centre line by 3.6 s. | verify-eclipse-here.mjs | Family H and the first pass's calculator run again: 1.47 km and 1.46 km outward; 3.6 s longer. - CONFIRMED | [auwers] The standard value of 959.63 arcseconds was determined by Auwers and published in 1891. | https://arxiv.org/pdf/2107.09416 | Kept text of the paper: "This is the value determined by Auwers in the late 1800’s (Auwers 1891) and it is used in all published eclipse predictions"; its reference list has "Auwers, A. 1891, AN, 128, 361". - CONFIRMED | [quaglia-radius] Quaglia, Irwin, Emmanouilidis and Pessi estimated the eclipse solar radius at the eclipse of 21 August 2017 as 959.95 arcseconds, give or take 0.05, at one astronomical unit. | https://arxiv.org/pdf/2107.09416 | Abstract: "The value of the eclipse solar radius during the 2017 August 21st total solar eclipse was estimated to be" (959.95 ± 0.05) arcseconds "at 1 au". Authors: Luca Quaglia, John Irwin, Konstantinos Emmanouilidis and Alessandro Pessi. - CONFIRMED | [worth-sun] Changing only the Sun's radius from 959.63 to 959.95 arcseconds moves each edge of the 2027 path inward by 0.55 km and shortens totality on the centre line by 1.4 s. | verify-eclipse-here.mjs | Family H and the first pass's calculator run again: 0.55 km inward at both edges; 1.4 s shorter. - CONFIRMED | [ign-page] Spain's Instituto Geográfico Nacional has a page for the eclipse of 2 August 2027 with a table of municipalities and an infographic for each. | https://eclipses.ign.es/eclipse-total-sol-de-2-de-agosto-2027.html | Kept copy: "En la tabla siguiente se indican los datos del eclipse para cada municipio." It has 8,131 rows, each linking its own infographic; its footer begins "Información astronómica oficial ofrecida por el Instituto Geográfico Nacional". - CONFIRMED | [ign-infographic] Its infographic for Casarabonela, in Málaga province, prints a longitude, a latitude and an altitude, 36° 47' 5.30" north and 4° 50' 35.15" west at 495 m, and 21 s of totality. | research/eclipse-here/sources/ign/29040_Casarabonela_Malaga.jpg | Read by eye again for this update: "Longitud: -4° 50' 35.15"", "Latitud: 36° 47' 05.30"", "Altitud: 495 m", "Totalidad: 0m 21s"; its SHA-256 is the one check/ign-2027.json records; the image is https://eclipses.ign.es/src/img/eclipse-27/infografia/29040_Casarabonela_Malaga.jpg. Since dff219b53b the note in check/ign-2027.json, which is served beside the page, says the same ("which prints a longitude, a latitude and an altitude"; it said "the coordinates and altitude it was computed for"), and so does the header of research/eclipse-here/ign-fit.mjs. MINOR as first read ("prints the point it was computed for"): finding 6. - CONFIRMED | [ign-row] The row for Casarabonela in the table on the same page puts the start of totality at 10:48:22 and its end at 10:48:32, official time: 10 s. | https://eclipses.ign.es/eclipse-total-sol-de-2-de-agosto-2027.html | Kept copy, the row for Casarabonela: 09:41:42, 10:48:22, 10:48:27, 10:48:32, 12:01:45; above the table, "Los tiempos corresponden a la hora oficial." - CONFIRMED | [casarabonela-32] At the printed point, the four switches at two settings each and the choice between sea level and the printed height make 32 combinations: two give no totality at all and the rest give from 7.3 s to 58.3 s. With the place kept at its printed height, the four switches alone give anything from no totality to 57.3 s. | verify-eclipse-here.mjs | Family I at c1a89dfe80 and again at dff219b53b, and the first pass's calculator run again over all 32: no totality in 2 (DE440 elements, NASA's lunar radii and the 959.95″ Sun, at 495 m, under either Delta T), shortest 7.34 s, longest 58.32 s (at sea level); at 495 m alone, none to 57.27 s. WRONG as first read ("one gives no totality at all and the rest give from 4.5 s"): finding 3. - OBSERVED | [read-by-eye] The official values were fetched on 15 November 2026, and the times on 19 infographics were read from the images by eye. | recreated: the 19 kept images under research/eclipse-here/sources/ign/ read again by eye, all 19 by the first pass and Casarabonela's again by this one | Every longitude, latitude, altitude, instant and printed duration in public/tools/eclipse-here/check/ign-2027.json is what its image prints; the 19 table rows in that file match the kept page cell for cell; the recorded SHA-256 of the page and of each image match the kept copies, which are dated 15 November 2026. The images have not changed since 510e66e973; that file has changed only in its note, at dff219b53b (the wording of finding 6), and its values and hashes are as they were. - CONFIRMED | [table-earlier] At first contact, maximum and last contact, in all 57 cases, the table's time is earlier than the infographic's, by 7 to 10 s and 8.8 s on average. | public/tools/eclipse-here/check/ign-2027.json | Recounted: 57 instants, the table earlier in every one, by 7 s once, 8 s 18 times, 9 s 27 times and 10 s 11 times; mean 8.84 s. - CONFIRMED | [two-fits] The infographics are reproduced by NASA's elements with a Delta T of 69.6 s, a solar radius of 959.95 arcseconds and the place at sea level; the table is reproduced by the same elements with a Delta T of 76.3 s, a solar radius of 960.00 arcseconds and the place at its printed height. | research/eclipse-here/ign-fit.mjs | The whole grid (two sets of elements, two lunar radii, three solar radii, two heights, Delta T from 68 to 80 s in tenths) was run again by the first pass with its own calculator: the same two settings come out best. Scored again here: 19 of 19 towns right, instants within 0.48 s on average (worst 1.60 s), and 19 of 19, 0.43 s (worst 1.31 s). - CONFIRMED | [nasa-polynomial] A polynomial NASA publishes for the years 2005 to 2050 gives Delta T as 62.92 + 0.32217 t + 0.005589 t squared, with t in years from 2000. | https://eclipse.gsfc.nasa.gov/SEhelp/deltatpoly2004.html | Kept copy: "Between years 2005 and 2050, calculate: ΔT = 62.92 + 0.32217 * t + 0.005589 * t^2 where: t = y - 2000". - CONFIRMED | [polynomial-2027] For August 2027 that polynomial gives 76.1 s. | verify-eclipse-here.mjs | By hand, with the page's own rule y = year + (month - 0.5)/12: t = 27.625 and 62.92 + 8.900 + 4.265 = 76.085 s. - CONFIRMED | [nasa-limb] NASA's path table for 2 August 2027 says: "The eclipse predictions presented here DO NOT include the effects of mountains and valleys along the edge of the Moon. Such corrections for the lunar limb profile may shift the limits ... by ~1-3 kilometers, and change the eclipse duration by ~1-3 seconds." | https://eclipse.gsfc.nasa.gov/SEpath/SEpath2001/SE2027Aug02Tpath.html | Kept copy, under Eclipse Predictions. The whole quotation as the page and the entry print it matches character for character once the markup is taken out (the cut here is this record's). Checked again at c1a89dfe80 and at dff219b53b. - CONFIRMED | [svs-smooth] NASA's Scientific Visualization Studio puts it this way: "Bessel's method for predicting eclipses pretends that the Moon is a smooth sphere, when in fact its terrain is more rugged and extreme than the Earth's." | https://svs.gsfc.nasa.gov/4517 | Kept copy of the page titled Umbra Shapes: the sentence is there word for word; checked again at c1a89dfe80 and at dff219b53b. - CONFIRMED | [true-limb-maps] The Besselian Elements team publishes maps for 2027 and 2028 that account for the Moon's limb and the terrain, and states their inputs: the JPL DE440 ephemeris, TT minus UT1 of 69.1840 s, a solar radius of 959.95 arcseconds and a mean lunar radius of 1738.091 km. | https://maps.besselianelements.com/map/help/TSE20270802/ | Kept copy: "The IQP method accounts for the topography on both the lunar limb and at the observer site"; its table gives JPL/DE440, TT minus UT1 of +69.1840 s, (959.95 ± 0.05) arcseconds at one au and 1738.091 km. The kept page for 22 July 2028 reads the same. It also says the map "will be updated a couple of months before the eclipse when accurate EOP predictions become available", which the page's row for these inputs repeats. - CONFIRMED | [centre-of-figure] NASA's account of its method says: "The center of figure of the Moon does not coincide exactly with its center of mass. To compensate for this property in their eclipse predictions, many of the national institutes employ an empirical correction to the center of mass position of the Moon. This correction is typically +0.50" in longitude and -0.25" in latitude." | https://eclipse.gsfc.nasa.gov/SEpath/ve82-predictions.html | Kept copy (research/eclipse-here/sources/web/nasa_SEpath_ve82-predictions.html), Fred Espenak's "Solar Eclipse Predictions with VSOP87 and ELP2000/82": the three sentences, character for character, in the entry and on the page, once the source's line breaks are read as spaces. "NASA's account of its method" is fair: NASA's page of elements for 2 August 2027 links the words "VSOP87/ELP2000-85" to this page. Not fetched again. Since dff219b53b the page's list of sources names this page (finding 19). - CONFIRMED | [arcsecond-km] By NASA's figure of 1.863 km to the arcsecond at the Moon's mean distance, those two shifts are about 0.9 km and 0.5 km. | https://eclipse.gsfc.nasa.gov/SEhelp/limb.html | Kept copy (research/eclipse-here/sources/web/nasa_SEhelp_limb.html), Espenak's "The Lunar Limb Profile and Eclipse Predictions": "1 arc-second equals 1.863 km at the Moon's mean distance of 384,400 km". 0.50 x 1.863 = 0.93 and 0.25 x 1.863 = 0.47; 384,400 km times one arcsecond is 1.8636 km. A shift of the Moon across the line to the Sun moves the shadow's axis on the fundamental plane by the same amount, to a quarter of a percent. Since dff219b53b the page's list of sources names this page too (finding 19). - CONFIRMED | [stephenville-measured] On 8 April 2024 that team measured 13.7 s of totality at Stephenville, Texas, near the northern edge of the path. | https://www.besselianelements.com/eclipse-maps-accuracy/ | Kept copy: "The Besselian Elements Team observed the eclipse from Stephenville City Park"; "This is what we measured experimentally: C2: 18h 39m 06.6s UTC C3: 18h 39m 20.3s UTC duration: 13.7 s"; the northern limit "bisects Stephenville, TX". - CONFIRMED | [stephenville-predictions] The predictions they compared for the spot ran from 65 s (timeanddate.com) through 57.4 s and 24.5 s (xjubier.free.fr, as read from its map and with lunar limb corrections) to their own 12.9 s. | https://www.iota-es.de/JOA/JOA2024_4.pdf | Kept text of the issue, Table 1 of Quaglia, Irwin and Pessi: 65, 62, 57.4, 36, 24.5 and 12.9 s; beside it: "the first prediction by xjubier.free.fr is read from the local circumstances displayed on the map, while its second prediction is computed by applying lunar limb corrections." - CONFIRMED | [check-nasa] From NASA's elements the engine finds the instant of greatest eclipse of all 22 eclipses within 0.05 s of the one NASA prints, and on the 1,263 rows of NASA's path tables where the Sun is at least 20 degrees up its limit points fall within 0.29 km of NASA's. | verify-eclipse-here.mjs | Run at c1a89dfe80 and again at dff219b53b: C1 0.0457 s, D3 0.2900 km. The 1,263 recounted here from the 16 kept tables. The first pass's calculator: the instant within 0.046 s for all 22, and NASA's 2,526 limit points on those rows graze the computed umbra (median 0.05 km off). - CONFIRMED | [check-second] A second computation in Python, which takes the apparent Sun and Moon from JPL's DE440 and uses no elements, agrees with the engine on elements fitted to the same ephemeris to 0.0012 s at first and last contact, 0.030 s at second and third, and 1.0 m at the edge of a path, over 490 places and 221 edge crossings. | verify-eclipse-here.mjs | Run at c1a89dfe80 and again at dff219b53b: F2 0.0012 s, F3 0.0303 s, F6 0.998 m; 490 places, 221 of 224 crossings. research/eclipse-here/de440_check.py --eph de440s.bsp, run again outside the repository, gives check/de440-check.json byte for byte. (As first read the two agreed as closely and both were wrong: finding 1.) - CONFIRMED | [check-third] A third computation, which asks Skyfield for the apparent Sun and Moon as each observer sees them and shares no geometry with the other two, agrees with the engine to 0.03 s at first and last contact and 0.04 s at second and third, over 113 places, and at the edge of a path to 14 m with the Sun at least 25 degrees up and 92 m with it as low as 11. | verify-eclipse-here.mjs | Run at c1a89dfe80 and again at dff219b53b: F9 0.0320 s, F10 0.0425 s, 113 places; F11 13.6 m over 28 crossings, F12 91.8 m over 4. research/eclipse-here/topocentric_check.py --eph de440s.bsp, run again outside the repository, gives check/topocentric-check.json byte for byte. A second way, without the engine: the first pass's Besselian calculator on the same elements, along the same 32 lines, is 13.6 m and 91.9 m from the third's crossings, and the 28 and the 4 split at 25° as said (the Sun at the middle of the two low paths 20.1° and 11.3° up). The edge figures are new since the first pass. - OBSERVED | [aberration-caught] The third exists because the first versions of the second computation and of the fitted elements both left out the aberration of the Sun, agreed with each other, and were both wrong by up to 0.6 km; a reviewer's check caught it before publication. | research/eclipse-here/apparent.py | The record: "A reviewer's check caught it on 2026-11-15", and the README's account. The first pass met the first versions and measured the same thing on its own: their axis stood 1.3 to 8.1 arcseconds from NASA's in declination and 18.7 to 22.4 in hour angle over the 22 eclipses, which is the annual aberration; a computation with the light time taken to the observer gave Casarabonela a totality 0.9 s shorter than either. The size, 20.5 arcseconds over one Earth radius, is 0.63 km with the Sun overhead. - CONFIRMED | [check-faults] Nine faults planted in a copy of the engine, a spherical Earth among them, are each caught. | verify-eclipse-here.mjs | Run at c1a89dfe80 and again at dff219b53b, family K: nine, each caught, the spherical Earth by 22 of the engine's checks and the others by 1 to 22. Six as first read, true then; the three added are bugs the engine reviews found (a grazing contact given up on, the walk to the edge setting off the wrong way, and the walk striding over a narrow path), each put back and caught. - CONFIRMED | [check-keys] The figures the page works out for itself, in its written tables and nearly all of its paragraphs, carry a key that the check recomputes and compares; in the tables its script fills in, the times in Universal Time and every row of the table of named sets are compared in a browser. | verify-eclipse-here.mjs | Entry and source record, the first half in the words this record proposed at c1a89dfe80 and the second, since a49593092d, in the words it proposed at dff219b53b (finding 17). The first half holds, as counted at dff219b53b; the page has not changed since. Counting numerals cell by cell (figures written in words, such as "about a kilometre and a half", carry no key anywhere): every figure of the page's own in its written tables carries a key, the only cells without one being the Stephenville table's seven durations, which are theirs ("The numbers are theirs"); and of the 25 blocks of prose that hold figures of the page's own, 21 carry a key on every one, the other four being the paragraph on the fits to Spain's page (the "of 19" of "19 of 19", twice), the paragraph on the third computation (0.6 km, about 20 arcseconds, 25°, 11° and 21 thousandths of a second unkeyed beside its seven keys), the paragraph on a fifth convention (0.9 and 0.5 km) and the note over the strip of named sets (25°); 21 blocks of 25, and 100 keyed figures against ten unkeyed, is nearly all. At dff219b53b, on a scratch copy of the served files, the check failed when a keyed figure in a sentence was changed (2 of 2: 58.3 s, and the 1.2 km of the new sentence on the elements) and passed when an unkeyed one was (Stephenville's 13.7 s, the strip's 25°). The second half now says what is compared. The page's script fills in two tables (public/tools/eclipse-here/app.mjs): renderContacts marks only the time in UT of each contact (data-out ut-c1, ut-c2, ut-max, ut-c3 and ut-c4), and renderRows marks, in every row of the table of named sets, the cells for what it says, when, and how far from the edge; family M loads the page at six addresses, compares every marked cell with what the modules return and fails on any marked cell it does not compute. The Sun's altitude and azimuth and the second clock are unmarked and not compared (planted at dff219b53b: the altitude one degree high or the second clock an hour late left the check passing, 118 of 118, while the when cell, the edge cell or the UT times altered failed it), and the sentence no longer says they are. Each row's name and description, which give its settings rather than a result, are not compared either; the sentence is about the figures the page works out, and every row's are. Run at a49593092d, in the repository and from an empty folder: "171 things a reader sees at 6 addresses (verdict, times, every row of the table) are what the modules return". The page's own sentence says the same at more length (keys-and-browser). MINOR at c1a89dfe80 and again at dff219b53b: finding 17. - CONFIRMED | [added: dek] One Spanish town gets anything from nothing to 57.3 s depending on how they are set, and Spain's own official page gives it two different answers. | src/content/strata/eclipse-here.md | The first pass's calculator, run again: Casarabonela at its printed height, no totality under 2 of the 16 settings and 57.27 s at most; the infographic prints 21 s and the table gives 10 s. The page's og:description and the card say "from no totality to 57 seconds". - CONFIRMED | [added: silent] Totality at one place, with the four assumptions most eclipse calculators settle for you as switches. | https://eclipse.gsfc.nasa.gov/SEbeselm/SEbeselm2001/SE2027Aug02Tbeselm.html | Dek. The page's description, og:description and twitter:description and the card's one-line summary now say "the four assumptions most calculators settle for you", and src/pages/llms.txt.ts "most calculators settle for the reader". Since dff219b53b the entry's plain line says it too ("the four assumptions most calculators settle for you"; it said "the four assumptions that calculators usually keep hidden"), and the narration's "quiet choices" are gone (narration). No "silent", "hidden" or "quiet" is left on any of the layer's surfaces (searched), and "every" stays only where it is qualified, "every calculator that treats the Moon as a smooth ball" (four-things). NASA's page prints three of the four and the map help pages all four, and both still settle them for the reader, so the words fit. MINOR as first read ("the four silent assumptions of every eclipse calculator"): finding 16. - CONFIRMED | [added: same-definitions] A second set of elements was fitted here, by the same definitions (the apparent places of both bodies), to JPL's DE440. | research/eclipse-here/de440_elements.py | Entry, and the page's first switch ("by the same definitions"). The program takes apparent places (light time, aberration and deflection) through apparent.py; run again here on de440s.bsp it gives elements-de440.json byte for byte. The refitted axis is within 0.08 and 0.37 arcseconds of NASA's in all 22 eclipses (first pass). WRONG as first read (places corrected for light time only): finding 1. - CONFIRMED | [added: every-eclipse] It holds 22 eclipses: every solar eclipse from 2027 to 2035, and two past ones kept as controls. | public/tools/eclipse-here/elements.json | Recomputed here from JPL DE440 alone (Skyfield): of the 114 new moons from December 2026 to January 2036, exactly 20 bring the Moon's penumbra onto the Earth between 2027 and 2035, on the page's 20 dates; the other two are 8 April 2024 and 12 August 2026. The types in the page's table are the ones on NASA's 22 pages. - CONFIRMED | [added: refusal] So within 3 km of the smooth-Moon edge, under your settings or under any of the named sets, the tool prints the smooth-Moon figure and gives no verdict. | public/tools/eclipse-here/state.mjs | Seen in a browser at c1a89dfe80: Casarabonela opens on "Too close to the edge to call" (33.3 s, 0.77 km inside). A place 3.74 km inside under NASA's page as published and 2.95 km inside under DE440 with the eclipse Sun is refused, with that reason given. With the IAU Moon switched in, Casarabonela, now 2.24 km inside, is still refused. Wellington on 9 March 2035, partial and 2.83 km outside the path, is refused. A place 10.2 km inside under its own switches and at least 9.4 km inside under every set gets "Total for 1 min 58.6 s". Seen again at dff219b53b: Casarabonela as before; Stephenville on 8 April 2024, 1.24 km inside under NASA's page as published, refused; the northern limit of 6 February 2027 in NASA's path table, 0.11 km outside, refused; Wellington as before. - OBSERVED | [added: not-asked] That is a description and not an attribution: other models can land on the same seconds, and nobody at the institute has been asked. | research/eclipse-here/README.md | The author's record: "Nobody at the institute was asked how either product was computed." The kept page and each infographic say whose the calculations are: the Observatorio Astronómico Nacional. - CONFIRMED | [added: published-combinations] Eclipse Here makes each one a switch, shows the same place under a table of named sets, and says when your settings give a published answer back. | public/tools/eclipse-here/conventions.mjs | Seen in a browser: under the answer, "These four switches match the named set “NASA’s page, as published”", and for a combination of the reader's own, "No named set in the table below uses exactly these four switches: this is your own combination". The table's introduction names the rows that give a published answer back (NASA's own, and for 2 August 2027 the two fitted to Spain's page). Since dff219b53b the JSON-LD description says it the same way ("a table of named sets shows the same place under each"; it said "each set of them is a named row"). MINOR as first read ("lists the published combinations as rows"): finding 9. - CONFIRMED | [added: commands] The second computation is python3 research/eclipse-here/de440_check.py --eph de440s.bsp, the third is python3 research/eclipse-here/topocentric_check.py --eph de440s.bsp | research/eclipse-here/de440_check.py | Entry and page. Run as written outside the repository, with JPL's de440s.bsp, each gives its published file byte for byte, as does python3 research/eclipse-here/de440_elements.py --eph de440s.bsp. The page now says that the second and the fitting program import apparent.py and research/where-the-shadow-fell/eclipse.py, and both are published under /checks/ (public/checks/research/, identical to the originals). MINOR as first read (no --eph, and an import that was not published): finding 15. - CONFIRMED | [added: check-folder] all of these are published under /checks/, and what they produce sits beside this page: elements-de440.json, and in check/ the other three (de440-check.json, topocentric-check.json, ign-fit.json), with the readings of Spain's page they start from (ign-2027.json) | public/tools/eclipse-here/index.html | Each file is there and each of the five links resolves to it: public/tools/eclipse-here/elements-de440.json, written by de440_elements.py, and check/de440-check.json, check/topocentric-check.json and check/ign-fit.json, written by de440_check.py, topocentric_check.py and ign-fit.mjs (each program's own output path; the second pass ran each again to give its file byte for byte). The readings are check/ign-2027.json, which ign-fit.mjs reads; the two Python checks take their places from check/check-cases.json, not from Spain's page, so the "they" of "they start from" is the fit, which is how a reader will take it. The programs, apparent.py and the other layer's eclipse.py are under public/checks/research/, identical to the originals. MINOR at c1a89dfe80 ("and the data they produce sits beside this page in check/", untrue of elements-de440.json): finding 20. - CONFIRMED | [added: relates] The second computation behind this tool reuses that layer's vector geometry, though not its Sun and Moon | research/eclipse-here/de440_check.py | The program imports the other layer's module for the observer and the discs and takes its Sun and Moon from apparent.py. Board note 3139ea holds the older module's omission. - CONFIRMED | [added: only-request] Apart from this site's own visit counter, that is the only request the page makes to anyone; the eclipse is computed in your browser. | worker.js | Loaded in a browser from a local copy with every outside request blocked and logged: at eight addresses with coordinates the page's own code asked nobody else for anything, and a link with q= asked nominatim.openstreetmap.org once (stopped here before it left the browser). As served, the site adds to its pages a same-origin visit ping (/api/pulse) and Cloudflare's analytics beacon (both on the live sibling /tools/sky-now/ as the first pass captured it on 15 November); worker.js spares only /tools/secret-santa/ the beacon. The human's standing ruling (board note 706d1d, 28 September 2026) is that Cloudflare is the host, not a third party, and such sentences stand, read through /privacy/. - CONFIRMED | [added: polynomial-2004] The table sits 6.7 s of Delta T away from them, near the 76.1 s that the polynomial NASA publishes for the years 2005 to 2050 gives for August 2027 | https://eclipse.gsfc.nasa.gov/SEhelp/deltatpoly2004.html | Kept copy: "Between years 2005 and 2050, calculate: ΔT = 62.92 + 0.32217 * t + 0.005589 * t^2"; for August 2027 it gives 76.085 s, and 76.3 less 69.6 is 6.7. WRONG as first read ("the polynomial NASA published in 2004"; the kept copy is dated 2007 March 27): finding 5. - CONFIRMED | [added: town-table] (the table of 19 towns: two official columns and four of the engine under named sets) | public/tools/eclipse-here/index.html | The first pass's calculator, run again: all 57 cells from NASA's elements as printed; the DE440 column against a direct computation on DE440 with the light time taken to each observer, 17 of 19 cells as printed and the other two within 0.1 s (El Bosque 33.4 against 33.5 s, Ronda 1 min 17.2 against 1 min 17.3 s). The two official columns are as the kept page and images give them. 3 of the nineteen (Chipiona, Rubite, Casarabonela) are total under some of the four sets and partial under others. Unchanged since the first pass. WRONG as first read (28 of the 38 cells then on DE440 off, Casarabonela 4.5 s where DE440 gives none): finding 4. - CONFIRMED | [added: fit-figures] totality in the right 19 of 19, every printed instant within 1.6 s and 0.5 s on average ... within 1.3 s and 0.4 s on average ... NASA's page as published gives neither product back: it is 3.2 s from the infographics on average and 6.5 s from the table | public/tools/eclipse-here/check/ign-fit.json | Scored again here with the first pass's calculator: 0.48 s (worst 1.60 s); 0.43 s (worst 1.31 s); NASA's page as published 3.20 s and 6.48 s, each with the place at sea level, as research/eclipse-here/ign-fit.mjs scores the named sets (at the printed height, 3.25 s and 6.17 s; either way it gives neither product back). The two official products agree on which 14 towns get totality, and the table's is the shorter in all 14 (recounted). - CONFIRMED | [added: narrower] IERS Bulletin A predicts a Delta T of 69.43 s. The table sits 6.7 s of Delta T away from them ... with that value in place of the fitted one, the table's instants come back within 0.5 s on average. | verify-eclipse-here.mjs | 76.3 less 69.6 is 6.7; the first pass's calculator with 76.085 s in place of 76.3 s: 0.47 s on average. - CONFIRMED | [added: switch-tables] (the two tables of what each switch is worth, and) 08:48:12 UT, when the path is 237.4 km wide and totality on the centre line lasts 4 min 54.3 s ... 188.1 km wide, 3 min 48.5 s | verify-eclipse-here.mjs | The first pass's calculator, run again at the page's reference points: every edge, width, duration and clock figure of all eight rows as printed (the elements' rows now 0.27 and 0.67 km, 1.0 and 0.6 s: finding 2). Centre-line totality 294.3 s and 228.5 s. NASA's own tables print 237 km and 4 min 53.8 s at 08:48 UT, and 189 km and 188 km at 04:00 and 04:02 UT. The widths are across the path, as NASA's are, not between the two limit points of one instant. - CONFIRMED | [added: sydney-clock] The second watches the path of 22 July 2028 as its centre line crosses the meridian of Sydney, 04:01:28 UT | verify-eclipse-here.mjs | This pass's own computation of the central line from Meeus's formulas, on NASA's elements: it reaches 151.2093° E at 04:01:27.714 UT, at 33.9009° S, and 4.8431° W at 08:48:12.136 UT, at 35.7218° N, as the page's two captions say. MINOR as first read (04:01:27, the clock cut short): finding 13. - CONFIRMED | [added: switch-prose] barely moves it ... more than half a kilometre ... about three seconds later ... about a kilometre and a half ... several seconds ... about half a kilometre ... a second and a half | verify-eclipse-here.mjs | Against the two tables: 0.03 and 0.02 km; 0.65 and 0.63 km; 3.0 and 3.4 s; 1.46 to 1.64 km; 3.6 and 3.9 s; 0.55 to 0.62 km; 1.4 and 1.5 s. The sentence on the elements that opened this line until dff219b53b ("by up to a second along its length and a few hundred metres sideways") now has a line of its own (elements-worth), and so does the comparison of the Moon's switch with NASA's 3 km (another-km). - CONFIRMED | [added: moon-switch] NASA uses a smaller Moon for the central phase than for the partial phases, to stand for the deepest valleys on its edge. The IAU's adopted mean is larger: it widens a path of totality and lengthens totality, and narrows a path of annularity and shortens the ring. | https://eclipse.gsfc.nasa.gov/SEmono/reference/radius.html | Kept copy: "we depart from standard convention by adopting the smaller value of k = 0.272281 for all umbral (interior) contacts" (the contacts of the central phase, total or annular; NASA's pages of elements give 0.272488 for the penumbra); "a smaller value of k is more representative of the deeper lunar valleys and hence the minimum solid disk radius"; "the smaller k produces shorter umbral durations and narrower paths for total eclipses when compared with calculations using the IAU value for k. Similarly, predictions using a smaller k result in longer umbral durations and wider paths for annular eclipses than do predictions using the IAU's k." On the engine at greatest eclipse of all 16 central eclipses, the IAU Moon widens each of the 9 paths of totality (the hybrid of 14 November 2031 is total there) by 2.9 to 14.7 km and lengthens totality by 3.0 to 5.4 s, and narrows each of the 7 paths of annularity by 3.0 to 8.4 km and shortens the ring by 4.2 to 6.3 s (6 February 2027: 3.02 km and 5.06 s; 9 May 2032: 8.42 km and 4.23 s). The Besselian calculator written for this update, with the shadow cones rebuilt for each radius, gives the same sign every time, in the duration (to 0.01 s) and in the shadow's radius at NASA's point of greatest eclipse. MINOR as first read at c1a89dfe80 ("for totality ... gives a wider path and longer totality", untrue of the annular eclipses) and passed then by this record: finding 25. - CONFIRMED | [added: sun-switch] The almanacs' value dates from 1891. Measurements made at recent eclipses give a slightly bigger Sun, which narrows every path of totality and widens every path of annularity. | https://arxiv.org/pdf/2107.09416 | Kept text: Auwers 1891, "used in all published eclipse predictions; for example in The Astronomical Almanac"; 959.95 from the 2017 eclipse; the kept 2024 article gives 959.9 to 960.0 from the contacts timed at Stephenville. The paper does not speak of the width of paths; the geometry does, and the engine at greatest eclipse of all 16 central eclipses bears it out: the 959.95″ Sun narrows each of the 9 paths of totality by 1.1 to 5.6 km and shortens totality by 1.1 to 2.2 s, and widens each of the 7 paths of annularity by 1.2 to 3.4 km and lengthens the ring by 1.7 to 2.6 s; the calculator written for this update agrees in every sign and to 0.01 s. "and widens every path of annularity" is new at dff219b53b. - CONFIRMED | [added: maps] the Besselian Elements team publishes one for 2 August 2027 and one for 22 July 2028, each drawn with two lines at every edge to show a band, not a line | https://maps.besselianelements.com/map/TSE20270802/ | Both map addresses the page links to answered HTTP 200 on 15 November 2026 to the first pass (the robots policy was asked first and allows them); not fetched again. Kept help page: "Two lines are presented at each path limit corresponding to two values of the eclipse solar radius (959.90" and 960.00") in order to display a region of uncertainty rather than a simple line." - CONFIRMED | [added: last-valley] The real one has a profile, and totality ends where the last valley on its edge lets sunlight through. | https://svs.gsfc.nasa.gov/4517 | Kept copy: "Each edge corresponds to a single valley on the lunar limb, the last (or first) spot on the limb that lets sunlight through." - CONFIRMED | [added: stephenville-table] (the table of seven durations for Stephenville, published) in the Journal for Occultation Astronomy ... A spread of fifty seconds for one spot, and the prediction that came within a second | https://www.iota-es.de/JOA/JOA2024_4.pdf | Kept text, Journal for Occultation Astronomy, Volume 14, No. 4, Table 1: measured 13.7 s; timeanddate.com 65; eclipse2024.org (#1) 62; xjubier.free.fr (#1) 57.4; eclipse2024.org (#2) 36; xjubier.free.fr (#2) 24.5; besselianelements.com 12.9. Beside it: "The first prediction by eclipse2024.org is taken from the local circumstances calculator, while their second prediction is approximately gathered from the eclipse simulator", which the page's rows "eclipse2024.org, its local circumstances calculator" (62 s) and "eclipse2024.org, read off its simulator by the team" (about 36 s) follow. 65 less 12.9 is 52.1 s; 13.7 less 12.9 is 0.8 s. The article's conclusion names what its own prediction used: "both the lunar and terrestrial topography, using an improved value for the eclipse solar radius". - CONFIRMED | [added: lede] that is why two careful sources can give one street different seconds of totality, or one of them none | https://www.iota-es.de/JOA/JOA2024_4.pdf | Different seconds: Spain's two products for Casarabonela, 21 s and 10 s. One of them none: the kept article reports an observer "just outside our predicted northern limit but inside all other predicted limits". The sentence before it now says every such calculator "has at least four things to fix". - CONFIRMED | [added: delta-t-table] (the table of 22 eclipses: Delta T on NASA's page, from the IERS, the gap, and how far it turns the equator) ... at 0.465 km for each second | https://maia.usno.navy.mil/ser7/ser7.dat | Recomputed here without the engine from the 22 kept NASA pages, the kept Bulletin and the kept measured values: every type, every NASA figure, every IERS figure and its basis, all 22 gaps (each the difference of the two figures as printed, and the same from the unrounded values) and every distance (the printed gap at 0.4651 km a second, the equator's speed as the Earth turns). Five distances (5 December 2029, 1 June 2030, 30 March 2033, 23 September 2033, 20 March 2034) differ by 0.1 km from the unrounded gap's product, as rounding allows. The gap widens from 1.4 s to 9.8 s without once narrowing. - CONFIRMED | [added: gap-2033] (the table of eclipses, the row for) 23 September 2033 (which reads 79.2, 70.5, extended, +8.7, 4.0) | https://maia.usno.navy.mil/ser7/ser7.dat | The Bulletin's formula gives 70.4527 s for that day: 79.2 less that is 8.747, and 79.2 less the printed 70.5 is 8.7. MINOR as first read (+8.8, from an extension rounded to 70.45 first): finding 12. - CONFIRMED | [added: measured-note] “Measured”, for a date already past, is interpolated between the U.S. Naval Observatory's monthly values of Delta T (the latest in the file is 69.21 s, for 1 October 2026). | https://maia.usno.navy.mil/ser7/deltat.data | Kept copy: its last line is "2026 10 1 69.2065". Interpolated by date here: 8 April 2024, between 69.1983 and 69.2018 s, gives 69.1991 (printed 69.20); 12 August 2026, between 69.1713 and 69.1816 s, gives 69.1750 (printed 69.17). “Predicted” is the Bulletin's daily table, 365 rows from 13 November 2026 to 12 November 2027: "a year ahead". - CONFIRMED | [added: nasa-details] for the 16 that have a path, the place within 0.11 km, the width of the path within 0.06 km and the duration within 0.21 s ... a median of 0.09 km ... a tenth of an arcminute, which is 0.18 km. The 5 rows that are more than 0.6 km off (the worst by 1.6 km) all have the Sun no more than 10° above the horizon | verify-eclipse-here.mjs | Run at c1a89dfe80 and again at dff219b53b: C4 0.1102 km, C5 0.0603 km, C6 0.2099 s; the five far rows have the Sun at 10° or lower, now among all 1,422 rows. A tenth of an arcminute of latitude at 36° is 0.185 km. The first pass's calculator: the duration at NASA's place of greatest eclipse within 0.21 s for all 16. - CONFIRMED | [added: nasa-rows] NASA also publishes each path as a table, a row every two minutes: 1,422 rows in 16 tables (19 of them, at the ends of paths, print only one limit). | research/eclipse-here/sources/nasa/ | Recounted here from the 16 kept path tables with a parser of this pass's own: 1,422 timed rows, 120 seconds apart, 1,403 printing both limits and 19 one, each of the 19 among the first or last rows of its path (7 of the 20 rows for 30 March 2033). Of the 1,422, 1,263 have the Sun at least 20° up, none of them with one limit. MINOR as first read (1,403, only the rows that print both limits): finding 11. - CONFIRMED | [added: shares-nothing] So there is now a third computation that shares no geometry with the other two. | research/eclipse-here/topocentric_check.py | The program asks Skyfield for the apparent Sun and Moon as each observer sees them: no fundamental plane, no Earth's centre. It shares what the page's bracket says it shares: the ephemeris, the library, the list of places and, to know when to look, the second computation's times (for the new edge crossings, its time of maximum at the middle of each path). MINOR as first read ("shares nothing"): finding 8. - CONFIRMED | [added: what-is-left] NASA's elements and the ones fitted to DE440 give contacts on the centre lines that differ by up to 1.3 s, and paths that lie up to 1.2 km to one side of each other while their widths differ by no more than 0.24 km. That is what the first switch is. This page does not say which set is nearer the truth. | verify-eclipse-here.mjs | Run at c1a89dfe80 and again at dff219b53b: 1.3 s, 1.2 km, 0.24 km over the 102 crossings with the Sun at least 15° up on the centre line; recomputed here from check/de440-check.json over all 110, 1.17 km sideways at most, and the widths within 0.30 km (that one with the Sun 11° up, in 2033). From the two files: at the middle instant the fitted x is 0.10 to 0.83 km behind NASA's in all 22 eclipses and y within 0.49 km, while the axis's direction and the umbra's radius agree to 0.4 arcseconds and 12 m (first pass). - CONFIRMED | [added: implied] The Sun's radius and the Moon's radius behind NASA's elements were read back out of the semi-diameters its pages print: 959.64″ and k = 0.272486 on average | verify-eclipse-here.mjs | By hand from the 22 kept pages (first pass): the Sun's printed semi-diameter times its distance (taken from DE440, which the sentence leaves out) averages 959.635 arcseconds, each page within 0.05; the Moon's semi-diameter over its parallax averages 0.272486. - CONFIRMED | [added: keys-and-browser] And the figures this page works out for itself, in its written tables and nearly all of its paragraphs, carry a key that the check recomputes and compares: 416 of them. Where a browser can be driven, the check also loads this page at six addresses and compares what a reader would see (the verdict, the times in Universal Time, and for each row of the table of named sets what it says, when, and how far from the edge) with what the engine returns. | verify-eclipse-here.mjs | Run at dff219b53b and again at a49593092d with a browser: exit 0, 122 passed, 0 failed, among them "416 keyed numbers on the page, each equal to the value computed above" and "171 things a reader sees at 6 addresses" (79 at c1a89dfe80, before the when and edge cells were compared). Family M compares exactly what the bracket lists: the verdict's headline and sub-line, the UT time of each contact, and each named set's says, when and edge cells. On a scratch copy holding only the served files, 118 passed, 0 failed; the when cell, the edge cell or the UT times altered fail it (3 of 3), the Sun's altitude or the second clock altered do not (2 of 2), which this sentence does not claim, nor, since a49593092d, the entry's (check-keys). The first half as counted in check-keys. (411 keys when first read, true then.) - CONFIRMED | [added: small-enough] elements.json (NASA's elements, as parsed, each with the address and SHA-256 of the page it came from), elements-de440.json (the set fitted here) and deltat.json | public/tools/eclipse-here/elements.json | The 22 hashes in elements.json and the 16 in check/nasa-paths.json match the kept NASA pages (first pass; neither data file's sources have changed); deltat.json carries the Bulletin's and the table's. The fitted file describes itself as "apparent geocentric Sun and Moon: light time and aberration". - CONFIRMED | [added: quoted-at-the-top] The limb statement quoted under “What no switch can do” is from the path table for 2 August 2027. | public/tools/eclipse-here/index.html | In the list of sources. The statement is the first quotation in that section; the box at the top paraphrases it and links there. MINOR as first read ("quoted at the top"): finding 14. - CONFIRMED | [added: sources-list] (the list of sources: names, titles, authors and dates, and since dff219b53b) The centre-of-figure correction NASA does not apply: NASA, “Solar Eclipse Predictions with VSOP87 and ELP2000/82”; 1.863 km to the arcsecond: NASA, “The Lunar Limb Profile and Eclipse Predictions”. | research/eclipse-here/sources/ | Right in everything it names, against the kept copies: "by Fred Espenak, GSFC Planetary Systems Laboratory"; the titles "Mean Lunar Radius", "Polynomial Expressions for Delta T" and "Umbra Shapes"; the Bulletin's date; the paper's title, four authors and year; the article's title, three authors and issue; the two books as NASA lists them at https://eclipse.gsfc.nasa.gov/SEpath/ve82-predictions.html. The new item's two titles are the kept pages' own (their title elements read "NASA - Solar Eclipse Predictions with VSOP87 and ELP2000/82" and "NASA - The Lunar Limb Profile and Eclipse Predictions"), linked at https://eclipse.gsfc.nasa.gov/SEpath/ve82-predictions.html and https://eclipse.gsfc.nasa.gov/SEhelp/limb.html, the addresses of the source record's centre-of-figure and arcsecond-km entries. Every quotation on the page now has its source in the list. MINOR at c1a89dfe80 (neither page named): finding 19. - CONFIRMED | [added: ut1] Times are UT1, which a clock keeping UTC shows to within a second. | https://maia.usno.navy.mil/ser7/ser7.dat | The engine subtracts Delta T from Terrestrial Time, which gives UT1, and the kept Bulletin predicts UT1 minus UTC between -0.07 and -0.25 s for the year ahead. Further out it holds by the rule that keeps UTC within 0.9 s of UT1 with leap seconds: on the page's own extended figures the difference would pass 0.9 s in 2031 unless a second is taken out, which has never yet been done. - CONFIRMED | [added: casarabonela] Casarabonela is a town in the hills of Málaga province, close to the northern edge of the path of 2 August 2027. ... This page opens there. | https://eclipses.ign.es/eclipse-total-sol-de-2-de-agosto-2027.html | Kept copy: the row reads Casarabonela, Málaga; its infographic prints an altitude of 495 m. Under NASA's page as published the place is 0.77 km inside the northern edge (the first pass's calculator, again), and the page's opening state is that longitude, latitude and altitude (public/tools/eclipse-here/state.mjs). - CONFIRMED | [added: address] The address holds the place, the eclipse and all four switches, so a link reopens on this exact answer. ... Anything it cannot read it refuses by name, in the answer box, and uses the default instead. | verify-eclipse-here.mjs | Run at c1a89dfe80 and again at dff219b53b: "6 of 6 states come back identical from their own URL" and, in a browser, each offered link reopens its state. Seen in a browser: an address with a date the page does not hold, a latitude of 95, k=0.3 and a part it does not know (foo) opens the default with four notes naming each. The tags a link carries to say where it was posted (ref, utm_ and the like) pass without a note, by design since a5dcdb2da8; they are known, not unread. Since dff219b53b such a link, like the bare address, keeps the page's own title (seen: ?ref=bsky-eclipse-here&utm_source=bsky), and an answer someone asked for gets its own. - CONFIRMED | [added: card-rows] The engine is checked against the 1,263 rows of NASA's published path tables where the Sun is at least 20° up, and against two other computations that use no elements at all. | src/data/tools.mjs | The Tools card. 1,263 recounted from the kept tables. The card's other figures (no totality to 57 seconds, 33.3 seconds, 0.77 km) hold, and its picture shows the answer box its alt text describes, with the two lines the page now adds under the verdict, as the page prints them. MINOR as first read ("checked against all 1,403 rows"): finding 10. - OBSERVED | [added: first-published] First published 16 November 2026. No corrections yet. | src/content/strata/eclipse-here.md | The entry is dated 2026-11-16 and the page's datePublished says the same; the page carries no Corrected line. Not yet published when this pass ended, and this pass made no request to the site, so the date holds if it goes up on 16 November. (15 November when first read, when the first pass found the page and the published check answering 404.) At dff219b53b the foot still reads so. - CONFIRMED | [added: fifth-convention] There is a fifth convention the tool does not switch. ... NASA chooses not to apply it, and neither set of elements here carries it. | https://eclipse.gsfc.nasa.gov/SEpath/ve82-predictions.html | Entry; the page says the same at more length. Kept copy: "We choose to ignore this convention and have performed all calculations using the Moon's center of mass position", and further down, "No correction has been made between the Moon's center of mass and center of figure." research/eclipse-here/de440_elements.py takes the Moon's centre from DE440 with no offset. The page adds that 0.9 and 0.5 km are "the size of what the switches on this page move" (they move an edge by 0.02 to 1.64 km in its two tables) and that the correction "alone can account for a difference like the ones above": fair. NASA also says it "has no practical impact on the present work" and is "of consequence only where careful timings are needed or for observations near the northern or southern path limits", which is where this page works. - CONFIRMED | [added: refraction] Refraction is left out as well, of the times and the edge as much as of the altitudes; it matters to them only with the Sun low. | public/tools/eclipse-here/eclipse.mjs | Entry and page. The engine says so: atmospheric refraction is among what it does not model, and its Sun's altitude is "geometric, no refraction". Refraction grows from about a minute of arc with the Sun 45° up to over half a degree at the horizon (standard figures, not sourced on the page), so "only with the Sun low" is fair. Since dff219b53b the one place it enters is the headline for a partial eclipse whose maximum comes with the Sun below the horizon, which takes sunset or sunrise as an almanac gives it, the Sun's centre 50′ below a flat horizon (34′ of refraction and 16′ of semi-diameter: ALMANAC_HORIZON in eclipse.mjs), and says so; the times, the edge and the altitudes are still geometric, as the sentence says (sunset; finding 18). - CONFIRMED | [added: classical-metres] Those metres are real and belong to the classical method, NASA's included: it takes the Sun and Moon as the Earth's centre sees them, and an observer is up to 21 thousandths of a second of light time nearer the Moon. So read this page's distances to the edge to a hundredth of a kilometre with the Sun high, and to a tenth with it low. | research/eclipse-here/topocentric_check.py | A second way, with the first pass's direct code on DE440, along the third computation's own lines: the edge found from the apparent places as the Earth's centre sees them, less the edge found from those the observer sees, is 91.1 m and -48.6 m on 30 March 2033 (the Sun 11.3° up) and -5.9 m on 2 August 2027 (81.8° up), against 91.9 m and -48.3 m between the first pass's Besselian calculator and the third. So the metres are the classical recipe's, not the engine's. 6,378 km at the speed of light is 0.0213 s. The note over the strip of named sets now says the same at 25° (strip-25; it said 15° at c1a89dfe80); the largest gap found with the Sun under 25° is 92 m. - CONFIRMED | [added: fit-heights] The fit cannot tell sea level from the printed height for the infographics: with the place at its printed height and a Delta T of 69.3 s they come back within 0.6 s on average, nearly as well. | public/tools/eclipse-here/check/ign-fit.json | The fit, which scores the printed instants: at sea level 0.48 s (worst 1.60 s), at the printed height with 69.3 s 0.55 s (worst 1.57 s), totality right in 19 of 19 both ways; the first pass's calculator gives the same. The printed durations lean further to sea level than the instants do: the computed totality is 0.74 s from the infographics' on average at sea level against 1.34 s at the printed height (ign-fit.json; 0.70 and 1.20 s here against the printed Totalidad). As written, about the fit, the sentence holds. - CONFIRMED | [added: sunset] (the answer for Moscow, 12 August 2026:) Partial at sunset: 17.2% of the Sun’s width covered as it sets. That is sunset as an almanac gives it: the Sun’s upper edge on a flat horizon, with the usual allowance for refraction (with no refraction, as the Sun’s centre reaches the horizon, 4.1%). The maximum, 84.0%, comes 41 minutes later, with the Sun below the horizon. | public/tools/eclipse-here/state.mjs | Seen in a browser at ?e=2026-08-12&lat=55.7558&lng=37.6173&place=Moscow (at a49593092d the headline and sub-line word for word as above), with "At sunset, as an almanac gives it, the Moon covers 17.2% of the Sun’s width and 8.4% of its disc" under the times and "partial, 17.2% at sunset" in the rows of the named sets (17.1% on DE440). Two ways without the engine. The Besselian calculator written for the third reading, on NASA's elements with NASA's Delta T of 71.4 s: the Sun's centre 50′ below the horizon at 17:12:38.2 UT, with 17.23% of the width and 8.43% of the disc covered; maximum at 17:53:12.2 UT, 84.00%, 40.6 minutes later; the centre on the horizon at 17:05:57.4 UT, 4.13%; first contact at 17:03:51.1 UT with the Sun 0.26° up, as the times table says. Skyfield on DE440, the Sun and Moon as the observer sees them, the same Delta T: 17:12:35.7 UT, 17.15% and 8.37%; maximum 17:53:12.0 UT, 84.03%; the centre on the horizon, 4.04%. (The first pass's calculator gave 17.2% at 17:12:38 UT, and its direct computation 17.1%.) A sunrise is right too, where c1a89dfe80's search went the wrong way: at 26° N, 53° W on 2 August 2027 the page now says "Partial at sunrise: 38.9% of the Sun’s width covered as it rises" and "The maximum, 99.97%, comes 32 minutes earlier", where c1a89dfe80's engine headlined the unseen maximum (99.97% "as the Sun comes up"); the calculator gives 38.91% at 08:58:23.7 UT, 32.0 minutes after a maximum of 99.97%. Every figure holds, and since a49593092d the parenthetical says what its figure is: the Sun's centre on the horizon with no refraction, 4.13% at 17:05:57.4 UT by the calculator above, and 4.05% at 17:05:55 UT by this reading's own computation on DE440 with Skyfield (the Sun and Moon as the observer sees them, NASA's Delta T of 71.4 s, the altitude without refraction), which also gives 17.14% with the centre 50′ down at 17:12:35.5 UT and 84.03% at maximum at 17:53:12.0 UT. (The upper edge, which the almanac's sunset follows, would with no refraction stay up until 17:08 UT, with 8.2% covered, as the third reading found and this one's computation gives, 8.17% at 17:08:01 UT; the sentence no longer puts the 4.1% there.) Where the Sun's centre is above the almanac's horizon even at maximum, the other sub-line changed at a49593092d reads, seen at 36° S, 153° W on 8 April 2024 under "Partial: 9.6% of the Sun’s width covered at most, with the Sun on the horizon": "At maximum the Sun’s centre is 0.8° below a flat horizon: by an almanac’s allowance for refraction it is still just in sight. With no refraction, 9.3% is covered as the Sun’s centre rises." The same computation on DE440 there, with NASA's Delta T of 70.6 s: the centre rises at 16:35:47.5 UT with 9.27% covered; the maximum, 9.55%, comes at 16:31:57 UT with the centre 0.766° down, above the almanac's 50′. MINOR as first read (the headline at the geometric horizon, 4.1% "as the Sun goes down"): finding 18. MINOR at dff219b53b ("with none, 4.1%", and "With no refraction at all, the most seen would be" 9.3%, where the upper edge with no refraction gives more): finding 22. Both fixed. - CONFIRMED | [added: local-date] Each clock column is headed by its own calendar date at maximum; a contact on another date says so. | public/tools/eclipse-here/app.mjs | Seen in a browser. Wellington, 9 March 2035, tz=Pacific/Auckland: "UT, 9 March 2035" over the UT column and "Pacific/Auckland (UT+13:00), 10 March 2035" over the local one (New Zealand keeps daylight time until April); contacts 20:35:52.6, 21:54:28.5 and 23:20:44.6 UT, and 09:35:52.6, 10:54:28.5 and 12:20:44.6 on the local clock, all on 10 March; the first pass's calculator gives the same instants. Papeete: last contact "01:07:02.7 on 10 March 2035" under "UT, 9 March 2035". Pitcairn: maximum after midnight UT, so "UT, 10 March 2035", and first contact "22:33:44.5 on 9 March 2035". Wellington's answer, "Too close to the edge to call", partial 2.83 km outside the annular path: the first pass's calculator with this pass's walk to the edge gives 2.83 km, and 3.13 to 5.33 km under the other five sets, where the page prints 3.12 to 5.33. Seen again at dff219b53b at the first pass's address for Wellington (lat=-41.2866&lng=174.7756): the same headings, instants and distances, the named sets now naming their Delta T ("as now extended", 70.7 s). - CONFIRMED | [added: answer-box] (the answer the page opens on:) These four switches match the named set “NASA’s page, as published”. Of the 9 named sets below, 8 give a central phase here (10.8 s to 57.3 s) and 1 does not. | public/tools/eclipse-here/conventions.mjs | Seen in a browser at c1a89dfe80 and again at dff219b53b, and in the card's picture. The first pass's calculator with this pass's walk to the edge, the nine sets at Casarabonela: 33.27 s and 0.77 km inside; 33.84 s, 0.80 km; 18.82 s, 0.25 km; 57.27 s, 2.27 km; 27.58 s, 0.53 km; partial, 0.02 km outside; 45.81 s, 1.45 km; 21.75 s, 0.33 km (at sea level); 10.75 s, 0.08 km. Every row of the table as the page prints it (10.8 s from the engine's 10.750009). Since dff219b53b three of the rows name their Delta T (set-labels), and the check compares each row's when and edge cells too (for NASA's page as published, 08:48:16.3 to 08:48:49.5 and 0.77 km inside). - CONFIRMED | [added: disc] At maximum, the right way up: straight up in the drawing is straight up in the sky as you face the Sun. | public/tools/eclipse-here/app.mjs | The code turns the Moon's position angle by the parallactic angle (the textbook formula). Checked against the sky on DE440 (Skyfield, altitude and azimuth): the drawn Moon's direction from the Sun's centre, counted from straight up, is -107.2° at Madrid on 2 August 2027 against -106.9° in the sky, -131.0° against -131.3° at Melbourne on 22 July 2028, and 129.0° against 127.7° at the Chatham Islands on 9 March 2035. - CONFIRMED | [added: form-notes] The coordinates are what is computed; a label is only a label, whoever gave it. A minus sign or a comma for the decimal point is read as you would expect. ... (a height above sea level is near enough) | public/tools/eclipse-here/state.mjs | Seen in a browser: lat=36,7 and lng=−4,8 (a comma, and the minus sign the page prints) are read as 36.7 and -4.8; a place= label is shown in quotation marks beside the coordinates and never replaces them, and beside q= it is refused ("place is a label and is never looked up"). The height box is above the ellipsoid; at Casarabonela 50 m more height takes 0.18 s from totality and 9 m from the distance to the edge, and 100 m 0.37 s and 17 m, so a height above sea level is near enough. Seen again at dff219b53b: lat=36,7&lng=−4,8 opens at 36.7000° N, 4.8000° W, and q=Valencia beside place=Malaga refuses the label by name (the look-up, stopped here before it left the browser, says it could not be reached). - CONFIRMED | [added: elements-worth] The elements move the whole path, by up to 1.3 s along its length and up to 1.2 km sideways over the paths this page holds, and barely change its width. | verify-eclipse-here.mjs | New at dff219b53b and keyed (drift.t and drift.latMax, the figures of what-is-left). Run at dff219b53b: on the 16 centre lines NASA's elements and the ones fitted to DE440 put the contacts up to 1.3 s apart (up to 2.8 s anywhere, in shallow partial eclipses); the path lies up to 1.2 km to one side over the 102 edge crossings with the Sun at least 15° up, all 16 paths among them, and over all 110, recomputed here from check/de440-check.json, at most 1.17 km, while the width changes by at most 0.24 km (0.30 km with the Sun 11° up in 2033), on paths from 31 to 781 km wide at greatest eclipse: barely. Changing the 1.2 in this sentence fails the check. MINOR as first read at c1a89dfe80 ("by up to a second along its length and a few hundred metres sideways", true at the two reference points and passed then by this record against them, but short of the page's own 1.3 s and 1.2 km over all the paths): finding 25. - CONFIRMED | [added: another-km] NASA says it of its own edge, drawn with its own small Moon, and on this page the Moon's radius is a switch that moves an edge by about another kilometre and a half. | verify-eclipse-here.mjs | The paragraph opens on NASA's statement for 2 August 2027, and there the switch moves the two edges outward by 1.47 and 1.46 km (the first table; 1.46 to 1.64 km in both). On the engine at greatest eclipse it changes the width by 2.9 to 3.6 km in 13 of the 16 central eclipses, 1.5 to 1.8 km an edge; with the Sun low or the path oblique it is more (12 August 2026, the Sun 26° up: 3.3 km an edge; 9 May 2032, 20°: 4.2 km; 30 March 2033, 11°: 7.4 km), as any shift across the shadow is where the shadow meets the ground at a slant. The sentence's point, that the refusal holds under every named set because the switch moves an edge as far as that, does not rest on the figure. It replaced "moves an edge by about half as much again" at dff219b53b. - CONFIRMED | [added: strip-25] With the Sun lower than about 25° at maximum, read these to a tenth of a kilometre. | verify-eclipse-here.mjs | The note over the strip of named sets; it said 15° at c1a89dfe80. The third computation's crossings against the engine, recounted here from check/topocentric-check.json with the Sun's altitude at maximum at each: 92 m at 7.0°, 48 m at 14.4°, 20 and 21 m at 19.7° and 20.1°, 14 m at 24.4°, 12 m at 27.1°, and no more than 10 m anywhere above 50°. So below about 25° the edge is good to tens of metres, not to the hundredth the page gives it with the Sun high, and the check's own split (F11 and F12) is at 25°, on the Sun at the middle of each path. - CONFIRMED | [added: set-labels] (three of the named sets, as labelled for 2 August 2027:) NASA’s elements, Delta T as now predicted, the IAU’s lunar radius ... elements fitted to JPL DE440, Delta T as now predicted, NASA’s radii ... DE440, Delta T as now predicted, the Sun as an eclipse measured it | public/tools/eclipse-here/conventions.mjs | Seen in a browser, in the table of named sets and in the strip, each with its description ("Delta T 69.428 s, one larger Moon for every contact (k = 0.2725076), the almanac Sun"; "the other set of elements, Delta T 69.428 s, NASA’s radii, the almanac Sun"; "DE440, Delta T 69.428 s, NASA’s lunar radii, a solar radius of 959.95″"). In the code each of the three takes the value deltaTFromIERS gives, whose basis the label names: "predicted" from the Bulletin's daily table for the two eclipses of 2027, "extended" by its formula after that (seen: 69.6 s for 22 July 2028, 70.7 s for 9 March 2035) and "measured" for the two past eclipses (seen: 69.175 s for 12 August 2026, 69.1991 s for 8 April 2024). MINOR as first read at c1a89dfe80 (their names and descriptions did not say that they took Delta T from the IERS), and passed then by this record: finding 25. - CONFIRMED | [added: limb-map] Before you choose where to stand, check a map computed from the Moon's real outline. The Besselian Elements team publishes one for this eclipse: 2 August 2027. ... (for any eclipse but 2 August 2027 and 22 July 2028:) This page links such a map only for 2 August 2027 and 22 July 2028. | public/tools/eclipse-here/app.mjs | New at dff219b53b, under the verdict within about 20 km of an edge under any named set, and since c3ad3a26e8 under the reader's own switches too (app.mjs line 123; finding 26); the second sentence as a49593092d words it (line 128, unchanged since). Seen at dff219b53b: Casarabonela, linking https://maps.besselianelements.com/map/TSE20270802/, and the northern limit of 22 July 2028 at Sydney's meridian, linking https://maps.besselianelements.com/map/TSE20280722/. Seen at a49593092d: Casarabonela as before, and the new sentence at Stephenville on 8 April 2024 (32.2207° N, 98.2023° W, refused, 1.24 km inside) and just inside the northern edges of 6 February 2027 (22.048° S, 40.98675° W, 0.09 km inside) and 26 January 2028 (0.94° N, 57.38793° W, 0.01 km inside), each refused. It is so: the page's links to maps computed from the real limb are the team's two, https://maps.besselianelements.com/map/TSE20270802/ and https://maps.besselianelements.com/map/TSE20280722/, with their help pages, and the script builds no other. The team's article on Stephenville (research/eclipse-here/sources/web/besselianelements_eclipse-maps-accuracy.html), which the page cites, has a figure comparing the northern limits predicted for that town in 2024, the team's real-limb one among them, but it is an article about a past eclipse, not a map to choose where to stand by, and the note no longer says that the page knows of no map for 2024 or 2026 (the kept copy's menu lists an "Eclipse Map 2026"). The two maps exist (maps, true-limb-maps). MINOR at dff219b53b ("This page knows of no such map for this eclipse to link to", shown at Stephenville above the page's own table of that team's real-limb prediction for the spot): finding 23. - CONFIRMED | [added: lede-opens] It opens on Casarabonela, a town in southern Spain where the settings decide whether there is any totality at all; put your own place in under the answer. | public/tools/eclipse-here/state.mjs | New in the lede at dff219b53b. DEFAULTS opens on Casarabonela at the point and height its infographic prints (seen at the bare address). The kept row of Spain's table puts it in Málaga province, which is in Andalusia. At its printed height the four switches give anything from no totality to 57.3 s, none under 2 of the 16 settings (casarabonela-32), and of the nine named sets 8 give it totality and 1 does not (answer-box), so the settings do decide whether there is any. The place search and the coordinate boxes come straight after the answer box. - CONFIRMED | [added: device-position] The address above and the link below now hold this position, to about a metre. Change the place before sharing the link if that matters. | public/tools/eclipse-here/app.mjs | New at dff219b53b. Seen in a browser given a device position of 37.123456789° N, 3.987654321° W: after "Use this device's location" the note reads so, and both the address bar and the share box hold lat=37.12346&lng=-3.98765, labelled "a device’s reported position" (the code keeps five decimals; the address keeps up to seven). Five decimals is 1.1 m of latitude and no more of longitude; here the rounding moved the place 0.36 m north and 0.38 m east. The share box is below the note: the note is in the answer box, the share box under the table of named sets. - CONFIRMED | [added: card-refuses] A verdict within 3 km of the edge of a path. The Moon’s real outline decides that and this computes a smooth Moon, so it prints the smooth-Moon figure, says it is too close to call, and for 2 August 2027 and 22 July 2028 points to the maps that use the real limb. | src/data/tools.mjs | The Tools card's refusal (line 145), with the dates as a49593092d gives them, unchanged since. The years are right (finding 24), seen at a49593092d: at Casarabonela the page refuses and links the map for 2 August 2027, and at 32.77° S on Sydney's meridian, 0.04 km inside the northern edge of 22 July 2028, it refuses and links the one for that eclipse; just inside the northern edges of the two annular eclipses of those years, 6 February 2027 and 26 January 2028, it refuses and says "This page links such a map only for 2 August 2027 and 22 July 2028" (seen; limb-map). The corner left at a49593092d is closed at c3ad3a26e8. The page refuses when the place is within 3 km of an edge under the reader's own switches or under some named set (verdict in state.mjs, given c.edge and c.nearest), and public/tools/eclipse-here/app.mjs line 123 now shows the note, with its pointer, when either is within 20 km, so the note follows both of the refusal's tests and comes with every "Too close to the edge to call". Seen at c3ad3a26e8, in a browser: at ?e=2028-07-22&lat=-32.37&lng=151.2093&dt=200 (a Delta T of 200 s, where the named sets use 69.2 to 72.1 s; the page accepts 0 to 200) the page says "Too close to the edge to call", 0.18 km inside, with every named set's edge 32.22 to 35.12 km away, and now shows the note ending "The Besselian Elements team publishes one for this eclipse: 22 July 2028", linked to https://maps.besselianelements.com/map/TSE20280722/; at ?e=2027-08-02&lat=36.59&lng=-4.8431&k=0.27&sun=965, refused as a partial eclipse 1.55 km outside the edge under those switches while every named set puts the place 21.67 to 23.96 km inside, it shows the note linked to https://maps.besselianelements.com/map/TSE20270802/; the bare address shows it with that link as before; and on Sydney's centre line, ?e=2028-07-22&lat=-33.8688&lng=151.2093, "Total for 3 min 48.5 s", 91.1 km inside, it stays hidden. With a49593092d's app.mjs served in its place, the first two refuse with no note and the other two are as now. MINOR at a49593092d (a refusal made by the reader's own switches alone came with no pointer): finding 26. - CONFIRMED | [added: need-not] Look it up in two careful places and you can get two answers: different seconds of totality, or totality from one and none from the other. Neither need have made a mistake. | src/content/strata/eclipse-here.md | Entry. It said "Neither has made a mistake" until dff219b53b, which claimed of every such pair what the page shows of some (finding 25). As it stands it says what the page shows: Spain's two official products give Casarabonela 21 s and 10 s, each given back by one setting of the switches (two-fits), and settings that each published source uses give it anything from none to 57.3 s (casarabonela-32). - CONFIRMED | [added: plain] A free tool that works out a solar eclipse for any place, for every eclipse from 2027 to 2035, and lets you change the four assumptions most calculators settle for you. It tells you when your settings match a published prediction. It gives no verdict within 3 km of the edge of a path, where the Moon's real outline decides. | src/content/strata/eclipse-here.md | The entry's plain line, rewritten at dff219b53b. Every solar eclipse from 2027 to 2035, and two past ones besides (every-eclipse); the four switches (four-things, silent); when the switches match NASA's page as published, or for 2 August 2027 either set fitted to Spain's page, the line under the answer names that set (published-combinations; seen), and for those the engine gives the published answer back (check-nasa, two-fits); no verdict within 3 km of the edge (refusal). The page also names a matching set that is nobody's prediction, and says so in the table's introduction; the line does not say otherwise. - CONFIRMED | [added: narration] That need not mean anyone has made a mistake. ... They do not all settle them alike, and some do not say which way they went. ... Here those choices are switches, and when your settings give back a published answer, the page says whose. | public/tv/lib/narration/batch-123.js | The three lines of the TV narration that changed at dff219b53b (they read "Nobody has made a mistake.", "... and many do not say which way they went." and "Here those quiet choices are switches, and the page names whose answer you have just rebuilt.": finding 25); the other four are as before and say what the page says. Some do not say: NASA's pages give their elements, Delta T and lunar radii but not the Sun's radius, which the page reads back out of their semi-diameters (implied), and Spain's page and its infographics give none of the four, saying only that the calculations are the Observatorio Astronómico Nacional's (kept copies). When the settings give back NASA's page as published or, for 2 August 2027, a set fitted to Spain's page, the line under the answer names it (seen). "need not": as need-not. - CONFIRMED | [added: llms] ?e=2027-08-02 (the eclipse, by the UT date of greatest eclipse) ... &place=Malaga (a label shown beside lat and lng; refused with q) | src/pages/llms.txt.ts | The guide for machines, two glosses new at dff219b53b. Each of the 22 dates the page holds is the UT date of NASA's greatest eclipse, and none of those instants is near enough midnight for any Delta T the page allows to move it across (elements.json: the nearest are 23:04:32.7 UT on 9 March 2035 and 01:55:25.4 UT on 2 September 2035); the answer box says "the UT date of greatest eclipse". state.mjs keeps place only beside lat and lng and otherwise refuses it by name; seen beside q=Valencia: "place is a label and is never looked up; it is shown only beside a lat and lng". - CONFIRMED | [added: run-anywhere] Run it yourself: node verify-eclipse-here.mjs. The file is published at /checks/verify-eclipse-here.mjs; in an empty folder it downloads this page, its modules and its data, and tests those. | public/checks/verify-eclipse-here.mjs | The published check at a49593092d, identical to verify-eclipse-here.mjs (SHA-256 beginning 2764b6d6e1e28524), copied alone into an empty folder and run with Node 22 as the page says, with https://artwaste.land sent to a local server over public/ by two modules loaded first, which also logged each request and would have refused any other (none was made). It asked for 14 files, the page (/tools/eclipse-here/), its four modules (app.mjs, eclipse.mjs, conventions.mjs, state.mjs) and nine data files (elements.json, elements-de440.json, deltat.json and six under check/), wrote them to a temporary folder (/tmp/eclipse-here-gvc8pe) and tested those: 118 passed, 0 failed, exit 0, in 39 s, as its header says ("about 40 s from an empty folder"). It passed every check it ran: family A's generator runs are skipped outside the repository, as the file says, and family M ran in a browser found beside Node, "171 things a reader sees at 6 addresses" and no script error. When it ended no /tmp/eclipse-here-* folder was left (there was none before), and the empty folder held only the check. With every download answered 404 it stops at the first ("FAIL could not download index.html: HTTP 404"), exit 1, and leaves no folder either. a49593092d moved const scratch = [] and its exit handler (now lines 81 and 82) above the download block (lines 87 to 96), in both copies, as this record proposed. Inside the repository the download block is skipped and the check passes, 122 of 122. WRONG at dff219b53b (exit 1 at once, "ReferenceError: Cannot access 'scratch' before initialization", before it downloaded anything): finding 21. Not yet published, so not tried against the site. ## Findings - [fixed] 1. HIGH (same-definitions, aberration-caught). The elements "fitted here, by the same definitions" were not: NASA's take the apparent places of the Sun and Moon, the first fitted set took places corrected for light time only, and the second computation was built the same way, so the two agreed to a thousandth of a second and both stood up to 0.63 km from DE440's own answer. The first pass found it from the files (the fitted axis 5.4 and 20.3 arcseconds from NASA's on 2 August 2027) and from a computation with the light time taken to the observer; another reviewer reported it first, and the author changed both programs at 22:44 UTC on 15 November to apparent places, refitted, and added a third computation. Re-read then and again at c1a89dfe80: the refitted axis is within 0.08 and 0.37 arcseconds of NASA's for all 22 eclipses, all three programs run again give the published files byte for byte, and the entry says "(the apparent places of both bodies)". - [fixed] 2. HIGH (worth-elements). "0.23 km south" and "0.56 km south" in the entry, the source record, the page's paragraph and both switch tables are now 0.27 km and 0.67 km, and the clock figures in those rows are 1.0 and 0.6 s where they were 0.5 and 0.3 s. The check section's "up to 1.0 s ... up to 1.1 km ... unchanged to 0.14 km" is now 1.3 s, 1.2 km and 0.24 km, and the page adds "This page does not say which set is nearer the truth." Confirmed again at c1a89dfe80 with the first pass's calculator, and unchanged at dff219b53b. - [fixed] 3. HIGH (casarabonela-32). "one gives no totality at all and the rest give from 4.5 s to 58.3 s" is now "two give no totality at all and the rest give from 7.3 s to 58.3 s" in the entry, the source record and the page, with the range at the printed height (none to 57.3 s) added; the dek and the card use that figure. - [fixed] 4. MED (town-table). The nineteen-town table was rebuilt from the refitted elements; its last column, the map-makers' inputs on a smooth Moon, was dropped, and "2 of the nineteen" became "3 of the nineteen". Unchanged since, and confirmed again cell by cell. - [fixed] 5. MED (polynomial-2004). "the polynomial NASA published in 2004" is now "the polynomial NASA publishes" on the page, and the check's own message no longer says 2004. The entry never gave the year. - [fixed] 6. MINOR (ign-infographic). "prints the point it was computed for" is now "prints a longitude, a latitude and an altitude" in the entry, the source record and the page, and the page adds that its fit cannot tell sea level from the printed height for the infographics (now a line of its own, fit-heights). - [fixed] 7. MINOR (check-keys). "Every figure the page works out for itself carries a key" is now "The figures in the page's tables, and nearly all of those in its paragraphs, carry a key" in the entry and the source record, and the page says "the figures this page worked out for itself, those in its tables and nearly all of those in its paragraphs"; the two fitted solar radii were given keys. The entry's new wording overstated in its turn: finding 17, corrected at a49593092d. - [fixed] 8. MINOR (shares-nothing). "shares nothing with the other two" is now "shares no geometry with the other two", followed by what it does share. - [fixed] 9. MINOR (published-combinations). "lists the published combinations as rows" (entry), "each published combination named" (search description), "each published combination is a named row" (card) and the page's headings now say named sets; the page's table is introduced with "Only some rows give a published answer back". - [fixed] 10. MINOR (card-rows). The Tools card's "checked against all 1,403 rows" is now "the 1,263 rows of NASA's published path tables where the Sun is at least 20° up". - [fixed] 11. MINOR (nasa-rows). Listed and left by the first pass. "a row every two minutes: 1,403 rows in 16 tables" counted only the rows that print both limits. The parser in research/eclipse-here/build-elements.mjs now keeps a row with one limit, and the page reads "1,422 rows in 16 tables (19 of them, at the ends of paths, print only one limit)". Recounted here from the 16 kept tables: 1,422 timed rows, 19 with one limit, every one among the first or last rows of its path. - [fixed] 12. MINOR (gap-2033). Listed and left by the first pass. The row for 23 September 2033 read +8.8; it now reads +8.7, the gap being taken between the two figures as printed. The last column is now that printed gap at 0.465 km a second, so five distances moved by 0.1 km (5 December 2029 1.3, 1 June 2030 3.3, 30 March 2033 4.0, 23 September 2033 4.0, 20 March 2034 4.1). All 22 rows recomputed here from the kept files. - [fixed] 13. MINOR (sydney-clock). Listed and left by the first pass. The page's 04:01:27 UT, the clock cut short, now reads 04:01:28 UT; this pass's own central-line computation puts the crossing at 04:01:27.714 UT. - [fixed] 14. MINOR (quoted-at-the-top). Listed and left by the first pass. The first line of the sources now reads "The limb statement quoted under “What no switch can do”", which is where it is quoted. - [fixed] 15. MINOR (commands). Listed and left by the first pass. The entry and the page now give the commands as python3 research/eclipse-here/de440_check.py --eph de440s.bsp and python3 research/eclipse-here/topocentric_check.py --eph de440s.bsp, and the page de440_elements.py with --eph too; the page says the three need Skyfield and de440s.bsp and that two import apparent.py and research/where-the-shadow-fell/eclipse.py, and the latter is now published under /checks/ beside the others (public/checks/research/where-the-shadow-fell/eclipse.py, identical to the original). Run here as written, each gives its published file byte for byte. - [fixed] 16. MINOR (silent). Listed and left by the first pass. "the four silent assumptions of every eclipse calculator" (dek), "the four things every calculator fixes silently" (description), "the four silent assumptions" (og:description), "the four silent assumptions of every calculator" (twitter:description) and "the four things every calculator fixes silently" (the card's one-line summary) now say "the four assumptions most (eclipse) calculators settle for you", and src/pages/llms.txt.ts "most calculators settle for the reader". No "silent" is left on the layer's surfaces. - [fixed] 17. MINOR (check-keys), new at c1a89dfe80, half right at dff219b53b, corrected at a49593092d. At c1a89dfe80 the entry and the source record said "The figures in the page's tables, and nearly all of those in its paragraphs, carry a key", untrue of the Stephenville table's unkeyed durations. At dff219b53b, in the words this record had proposed, they said "The figures the page works out for itself, in its written tables and nearly all of its paragraphs, carry a key that the check recomputes and compares; the tables its script fills in are compared in a browser": the first half right, the second loose, a mistake of this record's own proposal, since of the times table the check compares only the times in UT, and the Sun's altitude and azimuth and the second clock carry no key and are not compared (planted on a scratch copy, the altitude one degree high and the second clock an hour late each left the check passing, 118 of 118). Now, in the entry (src/content/strata/eclipse-here.md line 51) and the check-keys entry of assay/sources/eclipse-here.json (line 297), as this record proposed at dff219b53b: "...; in the tables its script fills in, the times in Universal Time and every row of the table of named sets are compared in a browser", which is what app.mjs marks and family M compares (check-keys). - [fixed] 18. MINOR (sunset), new at c1a89dfe80. Where the Sun set (or rose) during a partial eclipse before (or after) its maximum, the headline gave the coverage at the geometric horizon, the Sun's centre at 0°, "as the Sun goes down": at Moscow on 12 August 2026, 4.1%, where the almanacs' sunset comes 6.7 minutes later with 17% covered. At dff219b53b public/tools/eclipse-here/eclipse.mjs finds both crossings, the geometric one and the almanac's (ALMANAC_HORIZON, the Sun's centre 50′ below a flat horizon: 34′ of refraction and 16′ of semi-diameter), and public/tools/eclipse-here/state.mjs headlines the almanac's: "Partial at sunset: 17.2% of the Sun’s width covered as it sets", with the maximum, 84.0%, "41 minutes later". Confirmed two ways (sunset): 17.23% at 17:12:38.2 UT on NASA's elements, 17.15% on DE440. Re-reading that code the author also found that the search for a sunrise went the wrong way and reported the unseen maximum, a case this record had not tried: at 26° N, 53° W on 2 August 2027, c1a89dfe80 headlined 99.97% "as the Sun comes up", and dff219b53b says 38.9%, which the calculator written for this update gives at 08:58:24 UT. Where the Sun's centre is above the almanac's horizon even at maximum, the page now says the maximum itself is just in sight. The new parenthetical "(with none, 4.1%)" was loose in its turn: finding 22, corrected at a49593092d. - [fixed] 19. MINOR (sources-list), new at c1a89dfe80. The list of sources named neither page the paragraph on a fifth convention rests on. At dff219b53b it ends with "The centre-of-figure correction NASA does not apply: NASA, “Solar Eclipse Predictions with VSOP87 and ELP2000/82”; 1.863 km to the arcsecond: NASA, “The Lunar Limb Profile and Eclipse Predictions”", each title linked to its page (https://eclipse.gsfc.nasa.gov/SEpath/ve82-predictions.html and https://eclipse.gsfc.nasa.gov/SEhelp/limb.html) and each as the kept copy gives it. - [fixed] 20. MINOR (check-folder), new at c1a89dfe80. "and the data they produce sits beside this page in check/" was untrue of elements-de440.json. At dff219b53b: "and what they produce sits beside this page: elements-de440.json, and in check/ the other three (de440-check.json, topocentric-check.json, ign-fit.json), with the readings of Spain's page they start from (ign-2027.json)", each file linked, and each link to a file that is there. - [fixed] 21. HIGH (run-anywhere), new at dff219b53b, corrected at a49593092d. The page tells a reader that the published check, in an empty folder, downloads the page, its modules and its data and tests them; at dff219b53b it stopped at once with "ReferenceError: Cannot access 'scratch' before initialization" (in verify-eclipse-here.mjs and public/checks/verify-eclipse-here.mjs the download block called scratch.push(dl) at line 86, and const scratch = [] was declared at line 107), so the download path always threw, while inside the repository, where that block is skipped, every run passed. a49593092d moved `const scratch = [];` and its `process.on('exit', ...)` above the download block in both copies, which are identical, as this record proposed (now lines 81 and 82). The published copy, alone in an empty folder with the site's address answered from a local copy of public/, downloads its 14 files and passes every check it runs, 118 of 118, exit 0, in 39 s, and leaves no /tmp/eclipse-here-* folder; with every download answered 404 it exits 1 at the first and leaves none either (run-anywhere). Once the page is live, run it from an empty folder against the site itself, as the brief's reader-side check asks. - [fixed] 22. MINOR (sunset), new at dff219b53b, corrected at a49593092d. public/tools/eclipse-here/state.mjs line 293 gave the geometric figure as "(with none, 4.1%)", that is, with no allowance for refraction; but the 4.1% is the Sun's centre on the horizon, and with no refraction the upper edge, which is what that sentence's sunset is, stays up 2.1 minutes longer, with 8.2% covered at Moscow. Line 296, for a Sun whose centre is above the almanac's horizon at maximum, said the same of the same figure ("With no refraction at all, the most seen would be ..."). Now, as this record proposed: "(with no refraction, as the Sun’s centre reaches the horizon, 4.1%)" at Moscow, and "With no refraction, 9.3% is covered as the Sun’s centre rises" at 36° S, 153° W on 8 April 2024, both seen in a browser and both figures recomputed on DE440 (sunset). - [fixed] 23. MINOR (limb-map), new at dff219b53b, corrected at a49593092d. public/tools/eclipse-here/app.mjs line 127 said "This page knows of no such map for this eclipse to link to." for every eclipse but 2 August 2027 and 22 July 2028, the two past ones included, though for 8 April 2024 and 12 August 2026 the page's own kept source, the team's page on Stephenville that the page cites, shows the team's true-limb limit for 2024 and lists an "Eclipse Map 2026", and the note appeared at Stephenville itself. Now, as this record proposed: "This page links such a map only for 2 August 2027 and 22 July 2028.", seen at Stephenville and at the edges of 6 February 2027 and 26 January 2028; it is so (limb-map). - [fixed] 24. MINOR (card-refuses), new at dff219b53b, corrected at a49593092d. src/data/tools.mjs line 145, the card's refusal, said the tool "for 2027 and 2028 points to the maps that use the real limb", but it points to one only for 2 August 2027 and 22 July 2028, not for 6 February 2027 or 26 January 2028. Now, as this record proposed: "and for 2 August 2027 and 22 July 2028 points to the maps that use the real limb", which holds at the edges of all four eclipses of those years under the named sets. The same sentence has one more corner, which no reading had tried: finding 26. - [fixed] 25. MINOR (moon-switch, elements-worth, set-labels, need-not, narration), not reported by this record at c1a89dfe80 and found by the house-rules review after it; all corrected at dff219b53b and confirmed here. The Moon's switch said the IAU's mean "gives a wider path and longer totality", untrue of the 7 annular eclipses the page holds, which this record had confirmed against NASA's words about total eclipses; the elements moved a path "by up to a second along its length and a few hundred metres sideways", which this record had confirmed at the two reference points while the page's own figures over all the paths were 1.3 s and 1.2 km; the three named sets that take Delta T from the IERS did not say so in their names or descriptions; and the entry's "Neither has made a mistake" and the narration's "Nobody has made a mistake", "many do not say which way they went" and "the page names whose answer you have just rebuilt" said more than the page shows (this record did not then cover the narration). Now: "it widens a path of totality and lengthens totality, and narrows a path of annularity and shortens the ring"; "by up to 1.3 s along its length and up to 1.2 km sideways over the paths this page holds", keyed; "Delta T as now predicted" (or extended, or measured) in each of the three names, and the value in each description; "Neither need have made a mistake"; and the narration as in its line. - [fixed] 26. MINOR (card-refuses), found at a49593092d, in the page since dff219b53b, corrected at c3ad3a26e8. The Tools card says that within 3 km of the edge of a path the tool says it is too close to call and, for 2 August 2027 and 22 July 2028, points to the maps that use the real limb. At a49593092d the pointer (public/tools/eclipse-here/app.mjs line 122) was shown when the place was within 20 km of an edge under some named set (c.nearest), while the refusal also comes when it is within 3 km of the edge under the reader's own switches (c.edge); with switches far from every named set's the two parted: at ?e=2028-07-22&lat=-32.37&lng=151.2093&dt=200 the page refused, 0.18 km inside, with no pointer, every named set's edge being 32.2 to 35.1 km away. Now, as this record proposed, app.mjs line 123 reads `$('limb').hidden = !((c.nearest != null && c.nearest < 20) || (c.edge != null && Math.abs(c.edge) < 20));`, under a comment saying the note is "shown when the place is near an edge under any named set, or under the reader's own settings", and nothing else in public/ has changed. Seen in a browser (card-refuses): at that address the page refuses as before and now shows the note and its pointer to the map of 22 July 2028; at 36.59° N, 4.8431° W on 2 August 2027 with k=0.27 and sun=965, refused 1.55 km outside the edge under those switches while every named set puts the place 21.67 to 23.96 km inside, it shows the pointer to the map of 2 August 2027; the bare address shows the note and that pointer as before; and on Sydney's centre line on 22 July 2028 the note stays hidden. With a49593092d's app.mjs served in place of the committed one, the first two show no note. At the first address, the switches set on the page to NASA's named set (the place then 33.2 km outside the edge) hide the note, and a Delta T of 200 s typed back in shows it again. Family M compares neither the note nor whether it is shown, so neither copy of the check needed a change, and neither has changed.