Repository navigation
Keep a joint's housing empty and every pocket open with corner pockets - #79
Merged
Merged
Conversation
…open With Mounting base set to corner pockets, a joint cut from beneath where four cells meet (a wall key's recess, an H-clip's, a dovetail's or a puzzle's notch) was cut from the cells and not from the bosses, so the bosses stood in the whole housing and nothing could go in. And a joint housed in a floor (keys at 2.8 mm, the puzzle's lobes at 2.6) or an extra floor stood round the 2.6 mm bosses and sealed their pockets in it. Both built with Download on (#70). - cornerBosses: the bosses are built only on an open underside. Over a floor the plate is built as a solid floor builds it, its pockets cut into the floor and the floor grown to suit them. - The joint's cuts from beneath are taken out of the bosses after their pockets. A boss a housing reaches stops a BLOAT short of the piece's edge, the two either side of one housing are one solid, and the cut is moved NUDGE into its own boss, so the cut adds no edge used four times (an H-clip at 42 mm goes from 40 a piece to 31) and none used once. A boss no housing reaches is built as before, byte for byte. - mountLimits holds a boss's pocket out of a joint's cut, as #64 does a cell's: with the housing cut from the boss, wall keys and the dovetail reach a boss pocket up to about 39 mm and the H-clip to about 34.5. The audit builds each joint cut from beneath, puzzle tabs and an extra floor with corner bosses, each way a pocket is cut, and reads the housing and the pockets along vertical lines: all 18 fail on the base and pass here. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
The menu still reads "Corner pockets only" when a joint's floor or an extra floor means the plate is built with a solid floor, so Checks says so, and the magnet depth is held to a boss's 2.4 mm only where the bosses stand. The page tests cover the note, both ways round, and the joint's limit on a boss's pocket at 36.13 mm. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
A boss's pocket is a lottery of its own: default screws in one-cell pieces leave 3 to 6 open edges on a boss's underside at 22 of the 901 pitches from 34 to 60 mm that take them, in bosses no housing reaches, as on the base. A boss a housing reaches is cut a BLOAT short of the piece's edge, and that draws the lottery again at other pitches: four one-cell pieces with dovetails, whose notches reach the bosses from 34 to 36.4 mm, left 3 open edges in two pieces at 34.64 and 34.72 mm, where the base built them closed. Those bosses now have their pockets cut again (cutAgain) when the first cut comes out open, and no boss a housing reaches leaks at any of the 901 pitches. Bosses no housing reaches are built as before. The audit builds both pitches. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
A boss a housing reaches has the joint's cut moved a NUDGE along each axis, so its walls stand a hair from the cell's and no edge is used four times. But a cut that comes out open is taken again a NUDGE along a diagonal (cutAgain), the cell's and the boss's alike, and that could put the boss's walls back on the cell's: a wall puzzle key at 36.13 mm with screws and a margin had 73 edges used four times in one piece, against 26 on the base. Moved two NUDGEs, no try of either cut lands on the other, and that piece has 21. The audit builds it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
A boss's pocket is a lottery of its own, joint or no joint: default screws in a one-cell piece left 3 to 6 open edges on a boss's underside at 22 of the 901 pitches from 34 to 60 mm that take them, and a 9.1 mm magnet at 39.46 mm 12 a piece. A boss no housing reaches now has its pocket cut again (cutAgain) when the first cut comes out open, as a reached boss's already is. Only when open. At 13 more of those pitches a whole boss's first cut comes out closed with a face turned over; unfoldFinished lays that fold out again, and every one of those pieces was watertight and oriented, so the first cut is kept and the bytes with it. Retaking those too would change 13 one-cell pitches and 10 magnet designs that had nothing wrong. Against the head before this, design by design: - one-cell drawer, default screws: 22 of 901 build new bytes, all 22 open before (81 edges), all watertight and oriented after; 879 the same bytes. - four one-cell pieces with dovetails, and with H-clips: the same 22 pitches each, open before (306 and 324 edges), 0 after; 879 the same. - one-cell drawer, magnets 6 mm and at their cap, from beneath and above, every 0.02 mm from 34 to 46: 26 of 2064 new bytes, all open before, 0 after; 2038 the same. Two at 43.9 mm (a 4.9 mm magnet) stay open with 3 edges: every try comes out open too, so they keep their bytes. No design changed whose first cut was closed, and none has more edges used four times. The audit builds the 37.7 mm one-cell drawer and the 39.46 mm magnet plate. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
mountLimits measures a corner boss's pockets against the joint's cuts with #69's bores, gaps and reasons, a pocket from above measured from the boss's top. The page names a boss for what stops a pocket, and leaves out the hint about keys put in from above with corner pockets. Both sets of page cases kept, with #69's figure for the 20 mm magnet. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
A boss is its pocket's depth and 0.8 mm (a magnet) or 1.0 (a screw head) tall, up to 2.6. One a joint's cut from beneath reaches, with its top level with the cut's ceiling (a dovetail's notch 2.4 mm, an H-clip's recess 2.3, a wall key's 2.0), came out open there: a 1.6 mm magnet or a 1.4 mm head beside a dovetail left 4 to 13 open edges a piece, at every pitch the notch reaches a boss. 0.01 mm off the level either way, none did. Such a boss now stands MOUNT_LEVEL over the ceiling, with a pocket from above cut from that top. Only bosses a housing reaches, within 0.05 mm of its ceiling, change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Only sw.js conflicted; the pages are rebuilt from the merged sources. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
A wall puzzle key at 36.92 mm with screws (a 2 mm hole, a 7.1 x 0.8 mm head), four cells square, kept a joint cut with a face turned over in the pair of bosses by the front edge's key, in A2 and in B2: three folds on the bed a piece, no edge open, and Download on. Every one of cutAgain's tries came back the same. Now a boss's cut that every try leaves open or turned is taken again moved two NUDGEs the other way along one axis, then the other, then both, and the first that comes out clean is kept; if none does, the first stands. Here the first, along y, builds both pieces with no folds and the same four-use edges. A cut that comes out right is built exactly as before, and the audit builds the design and counts its folds. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
mountLimits measured the joint's cuts where they stand in the cells, and a corner boss's cut stands two NUDGEs along each axis off that, up to 4.8 microns nearer a pocket: a pocket at the cap was 0.0051 to 0.0055 mm short of an H-clip's recess in the boss and 0.0057 to 0.0070 of a key's in the walls, against the 0.01 it was meant to keep. In a boss a pocket now stops that much further off. At every 0.01 mm of pitch from 30 to 60 with the five joints that reach the bosses, that lowers 2,676 of the 33,430 caps a joint sets (8%), all by a tenth but 6 dovetail ones by a hundredth, and starts a dovetail's refused sizes a hundredth lower in 5,392 of its 24,008. Without corner bosses nothing moves, and no build changes: NUDGE moves out of buildPiece, the same 1.7 microns, for mountLimits to use. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
…bove With corner pockets over a key housed in the floor, a refused magnet or screw size went on to say that keys inside the walls, put in from above, keep out of the solid floor under the pockets. Such a key needs no floor, so taking the hint stood the bosses up again with the key's cup in their pockets, #75's case. The hint is now left out wherever corner pockets would stand without the joint's floor, and kept under an Extra floor, which keeps them buried as a solid base does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
The level boss's comment said 4 to 9 open edges where the dovetail left 4 to 13; it gives each joint's now: 4 to 13 beside a dovetail, 9 to 30 an H-clip, 4 to 8, 9 to 18 and 25 to 74 a bowtie, a snap clip and a puzzle key in the walls. ENGINE.md no longer lists a puzzle's notch among the cuts taken from bosses: the puzzle's lobes always have their floor. The two NUDGEs keep a boss's walls off the cell's along each axis while at most one of the two cuts is taken again, and the notes now say so, not more. And the level test is left unrounded, with the reason: rounded to the micron as mountLimits' is, the bosses raised now at exactly MOUNT_LEVEL (2.25 mm by an H-clip, 2.05 by a key in the walls, 2.45 by a dovetail) would not be, and their bytes would change, where both ways build clean. No build changes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
Oliver-Johnson
pushed a commit
that referenced
this pull request
Oct 9, 2026
Only sw.js conflicted, on its VERSION line: taken from main and rebuilt with node build.js, which regenerated both pages. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XjNf7SQF9o4ZVeWiSbWQmk
Oliver-Johnson
pushed a commit
that referenced
this pull request
Oct 9, 2026
mountLimits measures a corner boss's pockets against the joint's cuts now, with the seam 2·NUDGE·√2 wider in a boss; the shank is measured turned there as in the cells (reach), since the boss's fastener is the same fastenerCutter. NUDGE is top-level on main, so buildPiece's local one goes. sw.js and the pages rebuilt. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY
This was referenced Oct 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Requested by Oliver · project thread
What this changes
Closes #70.
Before. With Mounting base set to corner pockets:
All of these built with Download on, and Checks said nothing.
After.
How.
cornerBossesbuilds the bosses only on an open underside. With a floor under the sockets, the plate is the solid floor's, byte for byte.Left for later, both on main as well: corner pockets at 34.04 mm with a 2 mm screw hole leave open edges under the bosses, joint or no joint (#80); and a magnet beside a counterbore of nearly its size at one depth folds edges in the bosses (#77's class).
New top-level names:
bossHeight,jointFloor,cornerBossesandNUDGE(moved up from buildPiece) in src/core.js. None is on main (5da18ce) or in #65. This branch contains main 5da18ce (#69 and #78), merged in cad0145 and 62eeb1d.Checks
src/, not the generatedindex.html/bins/index.htmlnode build.jsand committed the regenerated pages alongside the sourcesnpm run test:allpasses (run here:npm test, ci-sim, and the plate UI specs, 170 passed; not the whole UI suite)If this touches geometry
node test/bin-audit.jsreportsall cases cleannode test/plate-audit.jsreportsall plates watertight, and no case has picked up bad edges that were not already named as known🤖 Generated with Claude Code
https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY