This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
Was the strength-3 result luck? Does it carry over to other slow airframes, and pass the fixed test sets any new default must pass?
What was tested
Nine wing runs on fresh layouts; the other three slow airframes on the crossing at three speeds. Then the adoption checks: a mixed test set (145 scenarios, normal gusts) and a storm test set (70 scenarios, storms plus a poor radio link), for three slow airframes and the idealised instant-response craft.
What was found then
Fresh layouts reproduced the result, and all nine airframe-and-speed crossing conditions improved; the slow-responding test craft fell from 119 to 8 lost per run at 22 m/s. But the adoption checks failed. In the storm test the wing's collisions rose from 308 to 750; the multirotor and eVTOL leaned the same way, not clearly. The mixed set showed no clear change, and the instant-response craft was identical.
What we learned later
Flying the two changes separately showed that removing the extra room, not early reaction, broke the storm results.
research/data/oos/tau_stormsplit.md
What changed
nothing
Corrections retained in the record
Earlier wording: So the stage's target sentence is met on this cell for the two shorter-lag aircraft
Record correction: The cycle's self-audit says the target had no registered test. The multirotor's 'flat' profile (0 / 0.7 / 0) is at the measurement floor, and the eVTOL's 8 m/s condition sits exactly on the 0.5 band edge.
research/reviews/researcher-c132.md items 1–3; research/data/oos/tau_presets_spd.md
Earlier wording: On a calm crossing the bubble was the problem — too big, causing the very conflicts it guarded against.
Record correction: Reviewers had already ruled at c126 that the bubble 'manufacturing conflicts' was inference, not an identified mechanism.
research/reviews/auditor-c126.md (F-TAU-MECHANISM)
Numbers and source references
- Nine
- fresh-seed wing runs (3 speeds × offsets 300 / 400 / 500)research/data/oos/tau_m2k3.md (a)
- 1 / 3 / 7
- wing lost per run on fresh seeds (1.3 / 3.3 / 7.3)research/data/oos/tau_m2k3.md (a)
- nine
- airframe-speed crossing conditions improvedresearch/data/oos/tau_presets_spd.md (Stage reading)
- 5 / 55 / 119 to 2 / 3 / 8
- delayed-sideways craft lost per run, today's vs charge (5.3 / 54.7 / 118.7 to 2 / 2.7 / 8)research/data/oos/tau_presets_spd.md
- 145
- scenarios in the matrix benchmarkresearch/data/oos/tau_adopt.md
- 70
- scenarios in the storm-and-bad-link setresearch/data/oos/tau_adopt.md
- 1 720
- scenario records (4 craft types × 2 settings × 215 scenarios)research/data/oos/tau_adopt.md
- 860
- paired comparisons (1 720 / 2)research/data/oos/tau_adopt.md
- 308 to 750
- wing storm-set collisions, today's vs candidateresearch/data/oos/tau_adopt.md
- 583 to 1 466
- wing storm-set craft lostresearch/data/oos/tau_adopt.md
- 67 to 146; 84 to 119
- multirotor and eVTOL storm-set collisions (unresolved)research/data/oos/tau_adopt.md
- 544 to 509
- wing matrix collisions (unresolved)research/data/oos/tau_adopt.md