The question
The previous entry found about one bad ending in ten at the 1 m/s disturbance package and a pad that stopped at 2 m/s, but said nothing about the winds between. How fast do bad endings grow there, does the station-keeping setting that helped at 1 m/s keep helping, and what does it cost on the approach leg, where the earlier world had almost no wall touches to lose?
The flight
The CRAFT campaign flew the same Seventh Avenue pad world on 200 fresh seeds: steady cross wind at 1, 1.25 and 1.5 m/s, gusts scaled with the wind and turbulence held, each flown as the hauler lands today and with the station-keeping setting (slT 0.75). The score is the 1,726 scheduled hauler landings per cell; arrivals held at start-up are reported separately. Vehicle response, loading and the disturbance model remain assumed. Shape, cut and cost readings were registered before the rows were flown.
What happened
Bad endings per 100 scheduled landings: 9.85 at 1 m/s (the previous 10.04, replicated on fresh seeds), 25.70 at 1.25 and 45.88 at 1.5. Each step added more than the one before (+15.9 then +20.2 points; registered shape: convex). With the setting: 1.39, 3.65 and 8.35, a material cut at every dose, mostly in losses rather than damaged exits. The pad still operated at 1.5 m/s: 1,369 of 2,000 haulers landed. On the approach leg the setting did not raise first wall touches for cruise entries; at 1.5 m/s it had fewer, 15 against 82 in 1,726 arrivals, lower on all 60 seeds that differed. The registered reading is that a cost was not shown, which is not the same as no cost. Arrivals the simulator holds at start-up and releases from a hover, an artefact outside the score, went the other way (82 to 102). Ten of eleven registered predictions hit.
What we learned later
Amendment (2026-10-10, CRAFT cycle 35, provisional). The wind package was then taken apart on 200 fresh seeds. The rise from 1.25 to 1.5 m/s is carried by the steady cross wind alone: with nothing random in it, it made no bad ending at 1.25 m/s and 24.96 per 100 at 1.5, about half of the full package’s rise above calm; gusts and turbulence added about 11 points each at 1.5 m/s (about 14 each at 1.25) and did not detectably reinforce each other. This entry’s 1.25 and 1.5 m/s cells replicated on the fresh seeds (27.87 and 47.14 per 100 against 25.70 and 45.88 here; with the setting 3.67 and 9.52 against 3.65 and 8.35). With the station-keeping setting the steady wind alone made no bad ending in 1,715 landings; most of what it left came with both random parts present and was mostly damaged exits. How well a real position hold would cancel a steady side wind is not established, so the steady share is a result for the modelled controller.
research/city/design/M-CRAFT-PAD-WIND-SPLIT-record.md
Follow-up (2026-10-11, CRAFT cycle 36, provisional). This entry’s curve and the later split rest on an assumed slow hauler response (1.5 s / 0.3 s). Under the steady part of the wind alone, the same hauler given the heavylift class’s quicker response made no bad landing in 1,720 at 1.5 m/s where the assumed body made 427, so the steady-wind part of the loss in this model belongs to that assumption, which has not been sourced for a real craft. The gusty package was not re-flown with the quicker body.
research/city/design/M-CRAFT-PAD-STEADY-LADDER-record.md
Follow-up (2026-10-11, CRAFT cycle 37, provisional). The CRAFT 36 follow-up’s ‘belongs to that assumption’ was narrowed: the quicker body changes the response and the planning tied to it, and under the steady wind either half alone removed the loss. Under the full gusty package the quicker body made 4.8 bad endings per 100 at 1.5 m/s against 47.2, with the planning carrying about 60 per cent of the cut. Both bodies are assumed.
research/city/design/M-CRAFT-PAD-BODY-FULL-record.md
What this does not establish
One simulated world, body, arrival rate and wind direction. The 1.25 and 1.5 m/s doses are model settings, not street-canyon statistics, and no real wind limit follows. Only the first wall touch per craft is recorded. Reviewers of the previous entry split on its soundness, and this entry’s proposed map sentence awaits the cycle-35 CRAFT review. No default changed: the setting remains a test key, and its adoption is a human decision.
The next question
Next on the track: splitting the 1 m/s package into steady wind, gusts and turbulence to name the component that carries the rise. That step closed on its research branch on 10 October 2026 and awaits integration and review; its result is not reported here until then. After that, the assumed hauler response time as a test cell, and the damaged-exit endings that remain under the setting.
Source references for reproduction
Optional repository references, pinned to this snapshot. Repository access may be required.