Repository navigation
Commit 530fed3
committed
Rails-shape the rung lift: 144-cell reasoning + 4x4 Morton cascade + qualia
Three operator steers, one arc:
1. THE 144 IS THE REASONING MATRIX (not a membership bit). #858 gated
rung lifts on flat is_perception_verb — the hand-rolling anti-pattern
one layer up from the one #842 fixed for relations. Now the cue
catalogue only licenses the that-complement (resemble is Mirrors but
takes no that-clause; Moore's cheap check); the REASONING is the cell
read: (family, tense) -> tense-modulated modal = the lift's epistemic
force. Epistemic lemmas minted into the matrix: SEE-class -> Mirrors
(modal 0.70), KNOW-class -> Abstracts (0.85) — WordNet independently
makes the same cut (verb.perception vs verb.cognition supersenses).
IRREGULAR_PASTS so knew/saw/understood classify with correct tense.
The catalogue<->matrix WELD test makes drift impossible. The matrix
earns its keep on the first text: 'the woman SAW that the tree was
good' (0.70) < 'they KNEW that they were naked' (0.85).
2. THE 4x4 MORTON CASCADE (as-is, not expanded; deterministic). One byte
per cell [fq:2|tq:2|fm:2|tm:2], high nibble = coarse quadrant =
nibble ancestry; 12x12 occupied in the 16x16 palette256 page,
reserves RESERVE-DON'T-RECLAIM. The table was secretly already
pyramid-shaped (base_prior's own superclass headers; tense_modifier
is class-shaped over t/3). Inverse-pyramid residual probe MEASURED +
pinned: mean 0.0774 (cells are small perturbations on quadrant
centroids — the pyramid claim holds on the mean); max 0.500 =
Grounds.lokal, with the outlier catalogue named (lokal is the axis
the carve compresses worst — input to the wordnet tune). Quadrant
centroids keep the class signal (graceful degradation asserted).
3. AWARENESS = BLIND x CONTEXT (quale = cell modal x Staunen-at-lift).
Scene-scale asserted: 3:7 knew = 0.85 x 0.182 = 0.155 vs 3:6 saw =
0.70 x 0.123 = 0.086 — both factors independently ordered (the
context factor alone must rise between them; the eat reversal lands
in between). Corpus-scale HONEST finding: field-mean staunen DILUTES
(1:4's early-arena entropy outranks 3:7's two-contradictions-among-
hundreds) — averages wash out events, rate-not-count one layer up.
Reflexivity stays the corpus crown; PROBE-QUALE-LOCAL (Delta-staunen)
pre-registered, never tuned in post hoc.
Queued probes logged, not built: PROBE-QUALE-LOCAL, PROBE-WORDNET-
QUADRANTS (families <-> WordNet 15 verb supersenses, in-house rail),
PROBE-COMMA-144 ((3x4)^2 mixed-radix comma drift — real mechanism, [S]
until a phase-reader is named; the golden-step null is the falsifier
template). O7 fence held: verb_table's 144 only, sigma_rosetta untouched.
Gates: 135 grammar tests green; clippy -D warnings clean (contract lib +
probe); fmt clean; fixture falsifier asserts green.
Board: EPIPHANIES E-RUNG-LIFT-RAILS-SHAPED-144-QUALIA-1, LATEST_STATE,
AGENT_LOG. Task #49.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki1 parent 4880e9f commit 530fed3
6 files changed
Lines changed: 622 additions & 27 deletions
File tree
- .claude/board
- crates
- lance-graph-contract/src/grammar
- lance-graph-planner/examples
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
| 8 | + | |
| 9 | + | |
1 | 10 | | |
2 | 11 | | |
3 | 12 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
| 8 | + | |
| 9 | + | |
| 10 | + | |
| 11 | + | |
| 12 | + | |
1 | 13 | | |
2 | 14 | | |
3 | 15 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
1 | 8 | | |
2 | 9 | | |
3 | 10 | | |
| |||
0 commit comments