Orbital element sets (TLEs) were pulled from CelesTrak's general-perturbations catalogue at build time and baked into this page — 721 objects across six populations. A published page can't reach the network, so instead of freezing positions I froze the elements: the six numbers that define an orbit. Your browser then integrates them forward to whatever moment you ask for, which is why this page keeps working, live, indefinitely.
Two-body Keplerian propagation: mean anomaly advanced by mean motion, Kepler's equation M = E − e·sin E solved by Newton–Raphson, then perifocal → equatorial rotation through argument of perigee, inclination and RAAN, and finally a Greenwich sidereal-time rotation into latitude/longitude.
This is not SGP4. Real tracking models atmospheric drag and Earth's oblateness (J2); this one models neither, so low-orbit positions drift from truth — small over hours, meaningful over days, especially for the lowest objects. Element sets carry epochs spanning —, and error grows with distance from epoch. Earth is treated as a sphere (R = 6,378.137 km) and latitude is geocentric, not geodetic. The globe's radial scale is logarithmically compressed so the 35,786 km geostationary belt fits beside 550 km Starlink — altitudes are emphatically not to linear scale. A few objects — Chandra, XMM-Newton and NASA's four MMS spacecraft — ride highly elliptical orbits swinging from ~10,000 km out past 170,000 km; in the census chart they're drawn as vertical perigee→apogee lines rather than points, and on the globe anything past ~2.45 compressed Earth-radii is dropped from the frame and reported as a count instead. Finally this is a sample, not a census: all stations, science, weather and GPS objects are included, but only 130 of 567 geostationary and 420 of 10,721 Starlink — enough to show the structure, not to count the sky. Treat it as a good-faith instrument, not a conjunction-assessment tool.