4 October 2026
Orbit studies on 4 October 2026
6 studies were read against their pass lines on 4 October 2026; 21 runs finished and 0 were still running at the end of the day.
The readings for each study follow, with the work still owed.
What ran and finished
The fuel gauge: the timing test (job 3432335) finished after 0.02 hours on a node with 128 CPU cores and 224 GB of memory.
The fuel gauge: the orbit fuel run (job 3432336) finished after 0.05 hours on a node with 128 CPU cores and 251 GB of memory.
The avoidance map: the timing test (job 3432514) finished after 0.02 hours on a node with 128 CPU cores and 251 GB of memory.
The avoidance map: the orbit avoidance prep run (job 3432515) finished after 0.05 hours on a node with 128 CPU cores and 251 GB of memory.
The avoidance map: the orbit avoidance screen 2 run (job 3432517) finished after 0.71 hours on a node with 128 CPU cores and 251 GB of memory.
The avoidance map: the orbit avoidance screen 1 run (job 3432516) finished after 0.71 hours on a node with 128 CPU cores and 251 GB of memory.
The reentry forecast: the timing test (job 3432257) finished after 0.04 hours on a node with 128 CPU cores, four A100 graphics processors and 503 GB of memory.
The intercept forecast: the timing test (job 3432622) finished after 0.1 hours on a node with 128 CPU cores and 251 GB of memory.
The reentry forecast: the orbit reentry record run (job 3432258) finished after 0.02 hours on a node with 128 CPU cores and 251 GB of memory.
The intercept forecast: the orbit intercept first run (job 3432623) finished after 0.57 hours on a node with 128 CPU cores and 224 GB of memory.
The reentry forecast: the training run (job 3432259) finished after 0.1 hours on a node with 128 CPU cores, four A100 graphics processors and 503 GB of memory.
The reentry forecast: the orbit reentry finish run (job 3432260) finished after 0.07 hours on a node with 128 CPU cores and 251 GB of memory.
The avoidance map: the orbit avoidance finish run (job 3432518) finished after 0.26 hours on a node with 128 CPU cores and 251 GB of memory.
The fuel gauge: the orbit fuel belt run (job 3438423) finished after 0.02 hours on a node with 128 CPU cores and 251 GB of memory.
The reentry forecast: the orbit reentry refit run (job 3438680) finished after 0.12 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the timing test (job 3444234) finished after 0.41 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the null test (job 3444237) finished after 4.93 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the orbit antenna planted run (job 3444236) finished after 4.72 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the orbit antenna fit run (job 3444235) finished after 3.06 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the orbit antenna curve run (job 3444238) finished within a minute on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the orbit antenna timing v2 run (job 3449102) finished after 0.42 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the orbit antenna null v2 run (job 3449105) finished after 4.7 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the orbit antenna planted v2 run (job 3449104) finished after 4.65 hours on a node with 128 CPU cores and 251 GB of memory.
The antenna study: the orbit antenna fit v2 run (job 3449103) finished after 3.05 hours on a node with 128 CPU cores and 251 GB of memory.
Failures and what was done
The reentry forecast: the first submission of the orbit reentry refit run (job 3438024) stopped with an error within a minute on a node with 128 CPU cores and 251 GB of memory. Cause: compares src/.code-sha with code-sha.txt and stops with one line naming when they differ, so a pushed tar that was never unpacked cannot run old code. The stand-in test covers the refusal. The fix: refuse to submit reentry jobs when the unpacked code is stale.
What the runs showed
The two full-depth screen runs (authority and deep): Met: 490 of 490 messaged passes found (line 97 percent); 959,029 passes under 1 km against 959,001 in the earlier record. Pass lines: 10 met, 1 not met.
The reentry forecast: Partly met, 3 of 9 lines: density error 0.79 of the model at 24 h; reentry median error 6.2 h against the authority's 3.3 h. Pass lines: 4 met, 5 not met.
The fuel gauge: Not met, 2 of 9 lines: planted 1 m/s burns found (GEO 100, LEO 92 percent); belt delta-v 17.8 times the passive floor, under 3 elsewhere. Pass lines: 2 met, 7 not met.
The avoidance map: Not met, 1 of 10 lines: placebo rate 0.82 of real (line one third); planned burns over 100 m found 21 percent; planted dodges 95 percent. Pass lines: 1 met, 9 not met.
The intercept forecast: Partly met: minute search finds 99.5 percent; belt recall 95 against 29 percent shuffled; predicted classes 0.97 to 1.31 times chance. Pass lines: 8 not yet read.
The antenna study: Not met: C22 449 and S22 44 bounds off the field; synthetic bias -14 and -7 bounds, scatter under one; real S22 falls to 3.2e-10. Pass lines: 1 partly met, 3 not met. This reading is provisional.
Next
The detector rerun: name the 15 objects behind the failed check against the published counts (11 passive, 4 payload) from an object table as of 20 September, and rerun the check with it. The close-pass study: a check of the low-orbit factor on debris or rocket bodies with precise orbits; today 11 objects (five laser spheres, six altimetry satellites) set every non-Starlink low ellipse, along-track factor 13.8. The burn-cadence detector: a declaration level chosen on held-out data that meets pass lines 1 and 2 together: thresholds rebuilt from each window's own resampled residuals, or a level set on the years the recall covers (the 2023 to 2025 passive windows give 80.959, where recall is 0.0302). The error atlas: rebuild the atlas files with factors per class and per era, split at 1 October 2025 (medium along-track 5.6 to 5.9 before, 3.39 after), from the finished fits with no refit, and rerun held-out coverage under them. The thermosphere study: the comparison with the published catalog-derived density record (Narayanan, Sicoli and Mehta, arXiv 2604.24646) and with Gondelach and Linares 2020: yearly quiet-day ratio and storm rise by shell. The two full-depth screen runs (authority and deep): the authority comparison outside the 37 days the public message store holds (needs a longer message archive). The reentry forecast: figure: reentry error against the number of sets in the fit window (what carries the two-day error). The fuel gauge: the belt delta-v record as a site reading (per fleet, per year), with the passive floor beside it. The avoidance map: a who-moves-for-whom reading on the fleets whose own ephemerides are public (SpaceX), with the ephemeris as truth instead of the catalog. The intercept forecast: a null matched to held belt satellites' slot spacing before the 3.5x excess is read. The antenna study: stage 0: the information bound for the two equatorial coefficients, the flattening term and its yearly change, computed from the error atlas before any model.
Built by code from the day's records; the house model's draft did not pass the checks. The sections on what ran and on failures are built by code from the facts. The day's facts ยท the check.