kerfmaster pin list · Big capabilities
recordStates what the code does today. Auditable against the code — this is the only type that can go stale.
Content last changed 2026-08-24 — computed from the item itself, not typed.
Jordan's ISV-H order (2026-08-21) was the first production-quantity nest: seven healed parts, fifteen each, mixed sheets. The first attempt composed a kit DXF carrying all 105 copies and re-healed it whole — 15 identical heals of each part, long enough to outlive its session twice. The fix is a qty parameter on split_parts.main: parts are replicated by shallow dict copy after plan_lead, safe because the packer and the y-first transpose only reassign part keys and never mutate the shared geometry lists in place. Copy ids append cNN to the base part id, so the manifest and any future manual-move file address each copy individually.
2026-08-24 -- material confirmed, eighth part added (ruling, Jordan): "Yes, let's stick with .080" / 14 gauge steel with .030" part tabs for this." Build 124 nests all 8 parts (his manual BNM78880 flat, build 123) at qty 15: 120 parts, 5 sheets, min part gap 0.2019, 0 held out. Sheet 1 carries the 105 small parts + 3 grille outlines (91.2in of X); sheets 2-5 carry 3 bnm76282 outlines each and leave a 45.3in usable remnant.
Posting the 124 sheets exposed two RECORDED GAPS (not hidden): (1) the post's chain walk re-judged nest joints at 0.008 when the nest stage stitched and coverage-gated them at Jordan's 0.020 closure ruling -- the -62/-67 second-tolerance defect again; fixed same commit (dxf_compat.chains takes tol, post_nest reads nest output at 0.02; TL-86 re-post regression byte-identical). (2) STILL OPEN: bnm76282's outer carries a micro-segment hairball (customer dirty CAD, spans 0.0003-0.002in); at the 0.02 walk ~10 fragments close into MICRO-LOOP boundaries that post as fake features -- sheet 2's trial post (build 126, O106) fails its own audit: comp wrong side on 5 boundaries, 16 comp-interference blocks, 27 tight pierces incl. clear-air 0.0000. THE AUDIT WORKING AS RULED (Jordan 2026-08-12): the program says do-not-run and none is delivered. Degenerate micro-loop boundaries need a principled drop (a closed loop shorter than the kerf cannot be a feature) before any ISV-H NC ships. Also open: the start/lead chooser is quadratic on the dense sheet 1 (108 parts, 6573 prims) -- one post exceeded 9m40s and was stopped; needs a spatial index before sheet 1 posts. The nest DXFs and MetaCam sheets are unaffected and delivered.