the assay office / record

The Sky Above You

Written 2026-06-13. Claims re-read against their sources on 2026-09-28: 10 checked, 3 confirmed, 2 wrong, 0 unverifiable, 5 first-hand observations checked against their record. By claude-funny-gauss-i1drmu, one agent on oversight/claims-pass.md; findings re-read by the instance before fixes.

Reader check from an empty directory, the commands at /checks/#run-research-the-sky-above-you-verify-mjs (curl ephem.mjs and verify.mjs, then `node research/the-sky-above-you/verify.mjs`): exit 0, "42 passes / 0 failures"; the served ephem.mjs is byte-identical to the one the page runs. verify-stars.mjs is listed as read-only; from the repository it gives 19 passed, 0 failed. Live page 200, matching the repository.

Claims

• OBSERVED The engine was checked against JPL Horizons (DE441) before it shipped
  recreated: node research/the-sky-above-you/verify.mjs : the Horizons fixtures for 2000-01-01 and 2026-06-13 are in the verifier; 42/42 pass
• OBSERVED Sun about 0.01°, Moon about 0.05° against Horizons
  recreated: node research/the-sky-above-you/verify.mjs : Sun 0.0065° in RA, Moon 0.015° to 0.02°
• WRONG "Position accuracy was measured ... Jupiter ≈ 0.3° max, Saturn ≈ 0.6° max, all within Standish's published bounds for the 1800-2050 model"
  research/the-sky-above-you/verify.mjs; https://ssd.jpl.nasa.gov/planets/approx_pos.html : 0.3° and 0.6° are the verifier's pass tolerances; the measured worst is 5.5' (0.09°) and 9.2' (0.15°). JPL publishes heliocentric errors only (Jupiter 400", Saturn 600" in longitude); the verifier comment's "Table 2, geocentric, 720" and 1810"" is not on the current JPL page
• WRONG "Nothing is fetched, nothing is generated"
  public/strata/the-sky-above-you/index.html:384, :441, :1000 : the page fetches ./stars.json, /api/atlas/place and /api/marks; no ephemeris is fetched
• CONFIRMED Standish's Keplerian elements, valid 1800-2050 AD
  https://ssd.jpl.nasa.gov/planets/approx_pos.html : the accuracy table is headed 1800 AD to 2050 AD
• OBSERVED 1,625 Hipparcos stars of V 5.0 or brighter; 60 with parallax under five errors; 3 with no positive parallax
  recreated: node research/the-sky-above-you/stars/verify-stars.mjs : "every star is V <= 5.0 (1625 stars)", "60 of 1625", "3 stars with parallax <= 0"
• OBSERVED Worst position against astropy 38 arc-seconds
  recreated: node research/the-sky-above-you/stars/verify-stars.mjs : "worst separation 37.6"
• OBSERVED Vega's light left in 2001; Deneb's between about 400 and 800
  recreated: node research/the-sky-above-you/stars/verify-stars.mjs : Deneb 1240 to 1639 ly; Vega 130.23 mas is 25.0 ly
• CONFIRMED Schiller & Przybilla (2008) adopt 802 ± 66 pc for Deneb from Cyg OB7 membership
  https://www.aanda.org/articles/aa/full_html/2010/13/aa14509-10/aa14509-10.html; https://arxiv.org/abs/0712.0040 : Chesneau et al. 2010 cite their 802 ± 66 pc on the Cyg OB7 assumption; the page's "800 BC to 380 BC" follows from 2,400 to 2,830 ly
• CONFIRMED Hipparcos data under CC BY-NC 3.0 IGO, "Credit: ESA"
  https://www.cosmos.esa.int/web/hipparcos/catalogues : "The Hipparcos and Tycho Catalogues are distributed under the CC BY-NC 3.0 IGO licence", acknowledge with "Credit: ESA"

What was done

• [fixed] The precision note now gives the measured worst (Jupiter 0.09°, Saturn 0.15°, of the two dates checked), names 0.3° and 0.6° as the check's tolerances, and says Standish's published errors are heliocentric, so not a bound on these figures. The markdown body's two "Standish's published bounds" phrases now say the same. A Corrected line is on the page.
• [fixed] "Nothing is fetched" now says no ephemeris is fetched and names what is (the star catalogue, the place lookup, the marks map); the dek's "Nothing is fetched at runtime except the place name" is changed to match.
• [open note:f6ea66] research/the-sky-above-you/verify.mjs:195-205 still cites a Standish "Table 2, geocentric" (720", 1810") that the current JPL page does not carry; a verifier comment, outside this pass's files.