kerfmaster pin list · Big capabilities
rulingmixedA decision by Aristide or Jordan. True because it was decided; it can be superseded, but it cannot be stale.
Content last changed 2026-08-24 — computed from the item itself, not typed.
Ruling (Jordan, 2026-08-24, from his marked-up screenshot of build 124 sheet 1): quoted in the hook. Measured facts behind each: every small part type on sheet 1 mixed 0/180 copies (p003: 14 at 0, one at 180 -- his circled anomaly); the 180s come from the packer flipping each COPY independently when it slides ≥0.1in tighter (the 2026-08-14 rotation gate scores inches only -- it now ALSO must not break PATTERN uniformity within a part type. Refined same day (Jordan): 'Push back on the P9997 hard rule: items like this can be 180 rotated in pairs.' His image: hook-shaped parts nested head-to-tail in alternating 180 pairs, interlocking. So the unit of uniformity is the repeating PATTERN, not the single orientation -- all-same and consistent A/B 180-pairing are both legal (a macro repeats a two-part unit as easily as one); a lone flipped copy among uniform mates remains the defect). The 172.5/352.5 tilts come from per-part min-area-rectangle rotation on slightly wedge-shaped outlines, blind to how tilted columns mesh. The sprawl comes from first-fit filling sheet 1 completely before opening sheet 2, with no remnant-shape term. P9997-style nesting macros (subprogram repeats of one part program at offsets) need uniform arrays -- an anomalous copy breaks the macro.
Record (built 2026-08-24, builds 129/130): rules 1 and 2 are IN THE PACKER: rotation variants are axis-aligned 0/180 only (per-part min-area tilts removed -- a tilt must buy measured packing to return), and a stray-suppression pass locks a part type to its majority angle and re-packs when under 25% of its copies sit at the minority angle; balanced mixes stand as legal 180 pairing. The candidate search scores one representative per part TYPE (O(n x types), not O(n^2)) -- qty 250 x 8 parts (2000 placed) packs in 20s. Nest material can now come from the Rev2 master (tables.rev2_row: lead_in + min_gap per row; Aluminum,0.080 = 0.125/0.25); stats carry material_source and a per-type orientation_mix census. Measured: 129-isvh-al15 (15 each, 48x120 aluminum) 3 sheets 58.3%, clean 65.3in cutoff on the last sheet; 130-isvh-al250 (250 each) 42 sheets 69.4%, min gap 0.2501 both. Rule 3 remains PARTIAL: consolidation falls out of y-first columns + biggest-first + the 0/180 restriction (129's layout shows one rectangular drop), but the remnant SHAPE is still not an explicit objective and later-sheet spare-band filling is not built.
Superseded intent text (kept for history): packer changes -- a regular orientation pattern per part type (all-same, or consistent 180 pairing when the shapes interlock and it buys packing; never a stray one-off flip), axis-align unless min-area tilt buys real measured packing, consolidate to low-X and prefer filling later sheets' spare bands so the drop is one rectangle. Re-run 48x96 and 48x120 ISV-H nests after, compare sheet counts before/after. MIXED flag: the three rules are his ruling; the packer redesign is intent.