kerfmaster pin list · The spec — what we build to
rulingA decision by Aristide or Jordan. True because it was decided; it can be superseded, but it cannot be stale.
Content last changed 2026-08-13 — computed from the item itself, not typed.
Contract — design stage
This contract states a decision, not an implementation.
takes —
makes —
fails if Any plate whose weight is carried by magnet friction; any magnet count justified against load rather than stabilisation; a finish or coating chosen by us; option A rebuilt with the cloud showing behind the circle before the incoming file is read.; any option C plate given a single pin without checking that a stack position exists above its CG; a pin placed at the CG rather than above it; option A rebuilt with the cloud showing behind the circle.
Contract last changed 2026-08-06 — computed, not typed. Dated separately from the text above, so neither date can speak for the other.
Jordan answered all seven open mounting questions on 2026-08-06. The load rule is the one with teeth.
“Unless we can create an environment where the shear factor is 10x the weight, provide a pin to carry the weight. The magnets are merely there to stabilize.” This supersedes the 4× bar cadmaster had been applying. At the recorded 2 lbf per magnet it decides every plate in the design without a further measurement: cloud 11.83 lb needs 118.3 lbf = 60 magnets, ring 2.44 lb needs 24.4 lbf = 13, triangle 1.05 lb needs 10.5 lbf = 6 (rounded up -- a partial magnet does not hold). None is reachable at three magnets a plate, so every plate gets a pin and no plate hangs on friction. The consequence worth keeping: the 2 lbf figure is no longer load-bearing anywhere in the design — it only has to be enough to stop a plate rattling — so paint, plating and finish stop being structural variables.
Raised by cadmaster, ruled by Jordan: “This only works if the pin can be set above the center of gravity. If not, we set a couple of pins and the rest is magnets.” A single round pin carries the weight but does not resist rotation about itself. With the pin directly above the plate's centre of gravity the plate hangs as a pendulum, gravity supplies the restoring torque, and the magnets only have to stop it rattling. Where no stack position sits above the CG, that plate gets a second pin running in a slot, for rotation only; the magnets are unchanged. This is a rule for option C and is applied per plate, never assumed.
Measured 2026-08-06 on the drawn geometry, requiring 0.40 in from the pin hole to any edge and 0.25 in of stem clearance past any plate behind it. Every plate has a usable position on the vertical through its own CG, so no plate needs a second pin on today's geometry: cloud body 8.00 lb, 0 to 4.44 in above the CG; top lobe 2.52 lb, 0 to 4.13 in; ring 2.43 lb, 5.68 to 6.65 in; triangle 1.05 lb, 0 to 2.30 in. The ring is the only plate that cannot take the pin at its CG — an annulus has no metal at its centre — so its pin sits at 12 o'clock on the band, which is also clear of the cloud behind it.
A consequence of the arithmetic rather than of the rule: a pin exactly at the CG is neutral, not stable. Restoring torque is W·d·sinθ, which is zero at d = 0, so the pin should sit as high above the CG as edge distance allows. On the cloud body at d = 4.44 in that is 0.62 lb·in per degree of rotation; at the CG it is nothing and the magnets carry all of it.
The plate list above already assumes option A's knockout, because it changes the count: cutting the cloud back around the circle severs the top lobe from the body, so option C becomes four plates, not three. The lobe was already hanging on a 0.002 in contact at the drawn ring position (recorded 2026-08-05), so the knockout reveals that weakness rather than causing it.
Jordan: “there's a big mistake in A — the cloud behind the circle should not show.” Aristide sent the agreed option A the same day (incoming/1786039753_3070743_file.jpg). Measured off it, scaled on the ring's 7.053 in OD: the cloud is cut back around the circle with a clearance gap — median 0.169 in, minimum 0.034 in, the ring touching the cloud nowhere — and the triangle is held to the ring by three bridges 121°/117°/121° apart, 0.22 in wide, each sitting just off a triangle corner. That is the pre-D16 geometry (our bridges were 0.202 in, inside the raster error). So D16's union is wrong here and D24's retirement of the bridges goes with it. The correct operation is neither the original subtraction (the ring's inner hole, no clearance) nor the union, but cloud.difference(outer_disc.buffer(clearance)) — a knockout by the element in front. Consequence flagged the same day: at that clearance the ring and triangle are attached to nothing, so option A cannot be one piece without bridges from the ring to the cloud as well.
Triangle joins ruled (Aristide, 2026-08-13): “follow my example” — the option-A rebuild reproduces the agreed photo’s pattern: three joins 0.22″ wide at 121°/117°/121°, each just off a triangle corner, never on the sharp point. No dots to come; the measured pattern IS the placement.