Skip to content

v0.506.16 — Scout 2026-11-10 sweep (2): diagnostic sky_falloff v3 samples a mosaic on one panel - #1432

Merged
JimmyeJones merged 3 commits into
mainfrom
claude/optimistic-cannon-fwsbuf
Nov 10, 2026
Merged

JimmyeJones merged 3 commits into
mainfrom
claude/optimistic-cannon-fwsbuf

Conversation

@JimmyeJones

Copy link
Copy Markdown
Owner

QA rotation sweep (2), new angle: the dark-ring gate diagnostic sky_falloff v2 pairs subs "within one target" so the object cancels — a single-field assumption. On a mosaic the sampled subs land on different panels, each framing a different part of the object, and the pairs MIN_PAIR_SKY_STEP admits are exactly those whose centre skies differ by the object. A vignette-free synthetic library two-thirds mosaics read 0.70 / 0.81 / 0.90 through run (bar: 0.97), which would falsely confirm the dark-ring mechanism.

  • sky_falloff.one_panel: cluster a target's solved centres at PANEL_LINK_DIST_DEG; a mosaic (≥2 panels or _mosaic name) is sampled on its fullest panel only; a _mosaic target with no located panel is skipped. Version 2 → 3 (v2 not yet on stable).
  • +5 tests in tests/test_diag_sky_falloff.py; the mosaic test fails before the fix (0.6983 / 0.8104 / 0.8961 vs 1.0 ± 0.02).
  • Docs: SHIPPED, IMPROVEMENTS (gate now names v3; Shipped line), FOCUS, PROCESS-NOTES.

Inbox: the same four owner-gated issues, unchanged since 2026-10-12. Dogfood --mosaic CLEAN. Full suite after syncing main: 8691 passed / 7 skipped.


Generated by Claude Code

…anel (sky_falloff.one_panel)

QA rotation sweep (2): v2's within-target pairs assumed one field. On a
mosaic the subs spread through its history land on different panels, each
framing a different part of the object, so the pairs MIN_PAIR_SKY_STEP admits
are the ones whose centre skies differ by the object -- and their slope is the
object's. A vignette-free synthetic library two-thirds mosaics read 0.70-0.90
through run(), under the dark-ring entry's 0.97 bar. v3 clusters a target's
solved centres at PANEL_LINK_DIST_DEG and samples only the fullest panel; a
_mosaic target with no located panel is skipped. Not yet on stable.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uj6RjzdgGcZoiCFAXRgXYm
…ut-fixed as v0.506.15 (FOCUS, PROCESS-NOTES, archive)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uj6RjzdgGcZoiCFAXRgXYm
…falloff v3 to v0.506.16 (webapp.__version__)

# Conflicts:
#	docs/FOCUS.md
#	docs/IMPROVEMENTS.md
#	docs/SHIPPED.md
@JimmyeJones
JimmyeJones merged commit 8672663 into main Nov 10, 2026
6 checks passed
@JimmyeJones
JimmyeJones deleted the claude/optimistic-cannon-fwsbuf branch November 10, 2026 23:09
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants