palette: spend the append margin on the 29 homeless TSV lanes; annotate the 4 overlaps

[AdaWorldAPI]

Rebased onto main after #1125.

What was measured first

contract::atoms::CANONICAL_ATOMS (the locked 33-lane TSV basis) and the 226-atom cognitive_palette had never been joined. Diffed by name: 0 exact matches, 4 accepted overlaps, 29 lanes with no palette home — and the 29 are not stragglers but four complete groups: all 9 Rung lanes, all 5 Σ, all 4 Presence, all 4 Meta knobs, plus 2/3 Pearl and 4/8 Ops.

The former append margin was exactly 29 slots (256 − (1 + 144 + 34 + 36 + 12)). Operator ruling: spend it.

What landed

• New Tsv block at 227..=255 (TSV_ATOMS, 29 entries), each carrying its canonical dim so the join back to CANONICAL_ATOMS is explicit. local is a palette address and deliberately diverges from dim, because the overlaps are skipped.
• TSV_OVERLAPS annotates the 4 lanes that already had a home rather than minting a second byte: abduct/deduce/induce → the ABDUCES/DEDUCES/INDUCES verbs (palette 85/83/84), transcend → ThinkingStyle::Transcendent (palette 213).
• AtomCatalogue::Reserved → Tsv(u8); ATOM_COUNT 226 → 255; RESERVED_BASE retained as the historical alias for TSV_BASE. No external consumers of either symbol.
• The palette is now FULL — const-asserted TSV_BASE + TSV_COUNT == 256, with a message saying no append margin remains.

The 29-fits-29 fit is coincidence, and the count is contingent

29 = 256 minus four block sizes on one side; 29 = 33 minus four overlaps on the other. The second number moves with the adjudication: 28 if imagine_counterfactual is read as the ICR recipe (rejected — a Pearl rung is not a named tactic), 30 if transcend/transcendent is read as distinct (that overflows 256), 33 on exact string match alone. TSV_OVERLAPS records each call with its target byte so the contingency is visible instead of buried in a constant.

Tests — 4 new, each disable-verified

• The partition law — every one of the 33 lanes is minted XOR annotated, exactly once. Falsifier run: pointing a TsvAtom at an overlapped dim fails with lane 18 covered 2x.
• Locals are dense AND diverge from dims for ≥ 12 lanes, so renumbering local to equal dim cannot pass silently.
• Every overlap resolves into Verb or Persona and points strictly below TSV_BASE. Falsifier run: aiming one inside the Tsv block fails.
• No byte value is unallocated; AtomId::tsv(29) is None, tsv(28) is 255.
• The existing 256-value census updated Reserved → Tsv.

Growth from here is ogar-loco, and the module doc says so

LaneShape::Pairs is 6 × (u8:u8) = function : value, so the 12-byte register is six Calls — a 12×u8 palette pick is a call of arity zero, the same ABI with every operand discarded. The Tsv block inherits that limit: rung_r1..r9 are nine needles for one op with a depth operand; the Σ chain and the Meta knobs are magnitudes stored categorically. A classid-selected vocabulary in loco's domain range has 112 slots per vocabulary against 29 shared globally.

The mechanism this points at is AdaWorldAPI/OGAR#295 (the shared basin codebook). This PR does not depend on it; the block is labelled the closing entry in a legacy table, not the pattern to repeat.

History note — the retraction pair is kept, by operator ruling

64462da5 added a probe, an atoms.rs note and a board entry reasoning about an I4x32 carrier against the V3 facet. The operator ruled it a hallucination. It had already been pushed, so 7ee96ace reverts it rather than force-dropping it, and the successor board entry records the correction in place.

Operator ruling: keep the pair, do not force-push. The branch stands as-is; the retraction is part of the record, not noise to be cleaned. (The one force-push this branch has seen was the rebase onto #1125, which was itself requested.)

Two further slips from that repair are recorded rather than hidden, because each earns a rule:

• git revert -q is not a flag. The invalid option made revert a no-op that printed usage, and a follow-up --amend then relabelled the original commit with the revert's message — producing a commit that added the hallucinated material under a "Revert" subject. The tell was git show --stat reporting insertions where a revert must report deletions.
• Rebasing onto W2b (D-DCR-2b): DN dissolution, mechanical/epistemic split, one-hop law — basin lanes + hydration primitives #1125 silently dropped one hunk of that revert. The revert originally deleted from three files; after the rebase it deleted from two, and the hallucinated board entry was live again on the branch. Main had prepended a new entry, so the hunk's context moved and the patch application dropped it — with the rebase reporting success and no conflict. Caught before pushing by asserting the reverted content is absent (git show HEAD:<file> | grep -c <marker> == 0) rather than trusting the exit code, and completed forward in 022f9f8f so the gap stays visible.

Verification

• cargo test -p lance-graph-contract — 1286 + 7 + 12 + 8 + 8 + 7 passing, post-rebase
• cargo clippy -p lance-graph-contract --all-targets --no-deps — clean
• Board: E-THE-PALETTE-MARGIN-IS-SPENT-AND-GROWTH-MOVES-TO-LOCO-1 prepended to EPIPHANIES.md

🤖 Generated with Claude Code

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_be71adcc-eade-4183-8660-f00ce60d111d)