Substrate-First — DRY gate + drift audit

The workspace — 18 repos carrying a root BOUNDARY.md as of 2026-09-01,
derived by the snippet in Step 2 rather than typed here — has a recurring
failure mode: an agent receives a task
("add threat perception"), jumps to implementation without checking existing
substrate, and builds a parallel system that duplicates functionality
already shipped under a different name. The user then has to catch it manually.

This skill prevents that. It runs in two modes:

1. Pre-implementation gate — run BEFORE writing code
2. Existing-code audit — scan for already-shipped DRY violations

Canonical failures (the pattern this skill prevents)

Failure 1 — ThreatField (Issue 047, riir-mmorpg-examples, 2026-08-01)

Built: ThreatField — a spatial hash grid (HashMap<(i32,i32), u32>)
for threat perception. Deposited monster positions into cells; NPCs sampled
the 3×3 neighborhood.

Already existed: GenericSpatialBelief<T> + target_within_visible_radius()

• decay_confidence() — the full fog-of-war → belief → decay pipeline in
  riir-games-shared/src/game_traits/spatial.rs.

Why the grep missed it: The agent searched for "threat field" /
"spatial hash". The substrate ships as GenericSpatialBelief /
SpatialBelief / confidence_decay. A single-vocabulary grep returns
ZERO hits even when the substrate fully exists.

Resolution: Reverted. The plain scan (tick_swarm_emotions) is a simpler
POC-scale simplification. The belief-based system is deferred until fog-of-war
becomes a gameplay feature.

Failure 2 — Orchard + Motivation in SDK src/ (Issue 490 + Issue 493)

Built: NpcReasonSystem, AppleGrowSystem, OrchardGoal, EmotionField,
EmotionAxis, tick_feeling_brain directly in the SDK facade's src/.

Already existed: The boundary rule (riir-game-sdk/AGENTS.md) says "no game
logic in src/" + Proposal 019 excludes emotion from the SDK. The substrate
belongs in riir-games (in riir-ai).

Why the grep missed it: The agent didn't check whether the types violated
the domain classification rule (latent semantic emotion ≠ raw physical
vocabulary).

Resolution: Extracted to riir-games (orchard) and riir-games::motivation
(emotion). The SDK re-exports them.

---

Mode 1: Pre-implementation gate (BEFORE writing code)

Run this checklist before implementing ANY of these:

• New impl System or tick function
• New trait + impl (perception, cognition, emotion, state management)
• New spatial query / index / hash / grid
• New vocabulary type / DTO / config struct
• New "helper" function that does math (distance, sigmoid, projection)
• New pipeline (perception → emotion → behavior, freeze → sync → thaw)

Step 1 — Vocabulary-translate your search

The concept you're building probably already exists under a different name.
Before grepping, write down 3+ name variants for the concept:

You're building...	Also search for...	Likely substrate names
"threat field" / "spatial hash"	belief, perception, spatial cognition, fog-of-war, visibility	GenericSpatialBelief, SpatialBelief, confidence_decay, target_within_visible_radius
"emotion" / "feeling" / "mood"	affect, drive, motivation, fear, desire	EmotionField, EmotionAxis, AffectField, DriveSpecSet, tick_feeling_brain
"state sync" / "delta" / "snapshot"	replication, gossip, cache, commitment	SyncBlock, ZoneDelta, PlayerStateCache, GossipDelta, SyncRegistry
"position" / "movement" / "physics"	spatial, coordinate, force, velocity	MapPos3D, ForceVector, SpatialIndex, GridSpatialIndex
"save" / "persist" / "freeze"	thaw, snapshot, serialize, store	freeze_avatar_delta, LocalKvStore, ShardIndex, NeuronShard
"validate" / "anti-cheat" / "check"	verify, guard, proof	AvatarAntiCheatValidator, AdaptiveModConfig, make_validator_predicate
"tick" / "update" / "loop"	system, schedule, game core	System, TickCtx, World, GameCore, FrameSnapshot
"knowledge" / "relationship" / "graph"	triple, semantic, KG	KgTriple, KgTripleTemplate, DualSignalEvidence
"attack" / "damage" / "combat"	hp, health, dex, fight	Hp, Dex, CombatConfig, combat_tick, attack_interval_ticks
"npc" / "swarm" / "crowd"	agent, bot, forager	SwarmState, ForagerSwarmSystem, ForagerAi, BotThought
"decay" / "fade" / "forget"	sigmoid, heal, baseline, confidence	decay_confidence, tick_feeling_brain, sigmoid, EmotionBaseline
"embedding" / "vector" / "latent"	direction, projection, HLA, shard	NeuronShard, HlaCacheProxy, compute_animal_emotions

This is the same technique as the paper→code vocabulary translation in
AGENTS.md §"Manifold Geometry". The R296 canonical failure applies internally
too: a concept-name grep returns zero hits because the math ships under operator
names.

Step 2 — Grep the CODEBASE (not just docs)

# Grep ALL repos for the substrate names from Step 1.
# Use multiple variants — the concept may exist under any of them.
#
# DERIVE the repo set; never type it (fixed 2026-09-01 — the hard-coded
# 7-repo brace list this replaced could not see 13 of the 20 contract repos,
# INCLUDING two product-set ones: riir-armageddon consumes
# `GenericSpatialBelief` in 2 files that the canonical DRY grep was structurally
# unable to find, and riir-dapps was equally invisible. A gate that cannot see a
# repo cannot tell you whether it consumes substrate or duplicates it.)
cd /Users/katopz/git
grep -rn 'GenericSpatialBelief\|SpatialBelief\|confidence_decay' \
    --include='*.rs' --exclude-dir=target --exclude-dir=.git \
    $(ls -d */ | while read -r d; do
        [ -f "$d/BOUNDARY.md" ] && [ -d "$d/.git" ] && printf '%s ' "$d"
      done)

Three deliberate details, each measured on this pattern (2026-09-01):

detail	why
-d "$d/.git", not -e	a git worktree has a .git file. ai-perfwt (a riir-ai worktree) otherwise contributes 43 duplicate hits — a DRY gate reporting one implementation twice is the failure it exists to prevent
--exclude-dir=target	the old form had none, and spent nearly all of its 23 s inside build dirs
unquoted $( … ), not $VAR	zsh does not word-split an unquoted parameter expansion (grep would get one giant argument and fail) but it does split an unquoted command substitution. Verified in this shell

Net: 227 hits in 1.4 s vs the old form's 221 in 23 s — strictly wider
coverage, no worktree duplicates, 16× faster.

# Also grep .research/ and .proposals/ for design rules that apply:
grep -rn 'two-brain\|fog-of-war\|domain classification\|sync boundary' \
    /Users/katopz/git/*/.{research,proposals,docs}/ 2>/dev/null

If you find existing substrate → STOP. Consume it. Do not build a parallel
system. Document why you're consuming it (in the plan/issue).

Step 3 — Check AGENTS.md architectural rules

Before building, verify your design doesn't violate these:

Rule	Source	What it means
Domain classification	AGENTS.md §"Latent vs Raw Space Rules"	Physical = raw exact; Semantic = latent dot-product + sigmoid; Social = KG triples
Two-brain model	AGENTS.md §"Spatial Cognition"	Info brain (synced ground truth) ≠ think brain (per-NPC beliefs, fog-of-war gated)
Sync boundary	AGENTS.md §"Sync Boundary Rule"	Through SyncBlock → quorum → Cold = raw + deterministic; Local = latent
Bridge pattern	AGENTS.md §"Bridge Pattern"	raw → latent = dot+sigmoid; latent → raw = clamp; zero-alloc, gateable
KG triple emission	AGENTS.md §"KG Triple Emission"	Semantic encounters → KG triple; Physical events → TxDelta with raw values
Facade constraint	riir-game-sdk/AGENTS.md	SDK = re-export facade, no engine deps; vocabulary in riir-games-shared
Boundary rule	riir-game-sdk/AGENTS.md	No game logic in consumer src/; game systems in riir-games

If your design violates any of these → STOP. File an issue. Rethink.

Step 4 — Decide: consume vs. build

Situation	Action
Substrate EXISTS and fits	Consume it. Wire via trait/config. Zero new substrate code.
Substrate EXISTS but wrong shape	Extend the substrate (in the right repo). File a plan.
Substrate DOESN'T exist	File an issue in the right repo FIRST. Then build.
You're not sure	STOP and file an issue. Don't guess.

Never build new substrate inside a consumer. The consumer provides data +
wiring only. If you're writing loops/math/constants in a consumer's src/ →
you're building substrate in the wrong place.

Step 5 — Record the decision

In the plan/issue, document:

## Substrate check (substrate-first skill)
- Searched for: [concept names + variants]
- Found: [existing substrate or "none"]
- Decision: [consume / extend / build new]
- Architectural rules checked: [list which rules apply + verdict]

---

Mode 2: Existing-code audit (scan for drift)

Run this when reviewing code, when you suspect a parallel system, or quarterly
as a DRY-hygiene gate (alongside boundary-guard).

Audit Step 1 — Inventory substrate primitives

For each domain, identify what substrate exists:

# Perception / spatial cognition
grep -rn 'GenericSpatialBelief\|SpatialBelief\|target_within_visible_radius' \
    --include='*.rs' /Users/katopz/git/*/  | grep -v '/tests/' | grep -v '/target/'

# Emotion / affect
grep -rn 'EmotionField\|EmotionAxis\|tick_feeling_brain\|AffectField\|DriveSpecSet' \
    --include='*.rs' /Users/katopz/git/*/

# State sync
grep -rn 'SyncBlock\|ZoneDelta\|PlayerStateCache\|GossipDelta\|SyncRegistry' \
    --include='*.rs' /Users/katopz/git/*/

# Spatial
grep -rn 'SpatialIndex\|GridSpatialIndex\|OctreeSpatialIndex\|MapPos3D' \
    --include='*.rs' /Users/katopz/git/*/

Audit Step 2 — Grep for parallel systems

For each substrate primitive found in Step 1, grep consumer code for
reimplemented versions.

Derive the consumer set — this block hard-coded
{riir-mmorpg-examples,riir-game-sdk} until 2026-09-01 and so could see 2 of
the 20 contract repos. That is the same defect the vocabulary step above was
fixed for, in the same file, one section down: 60655c48 corrected the Step 2
named "Step 2" and left this one, which is the step that actually looks for
duplicate implementations. riir-armageddon — a product-set repo that consumes
GenericSpatialBelief in 2 files (6 sites) — was invisible to every grep below.

cd /Users/katopz/git
consumers() {
  ls -d */ | while read -r d; do
    [ -f "$d/BOUNDARY.md" ] && [ -d "$d/.git" ] && printf '%s ' "$d"
  done
}
# Look for inline distance math (should use MapPos3D methods):
grep -rn '(dx.*dx.*dy.*dy).*sqrt\|distance_2d.*fn\|fn.*distance' \
    --include='*.rs' --exclude-dir=target --exclude-dir=.git $(consumers)

# Look for inline sigmoid/exp (should use substrate sigmoid or tick_feeling_brain):
grep -rn '1\.0\s*/\s*(1\.0\s*\+\|sigmoid\|exp(' \
    --include='*.rs' --exclude-dir=target --exclude-dir=.git $(consumers)

# Look for HashMap-based spatial structures (should use SpatialIndex substrate):
grep -rn 'HashMap.*i32.*i32\|spatial.*hash\|cell.*grid' \
    --include='*.rs' --exclude-dir=target --exclude-dir=.git $(consumers)

# Look for parallel belief/perception types (should use GenericSpatialBelief):
grep -rn 'struct.*Belief\|struct.*Perception\|struct.*Visibility\|last_known' \
    --include='*.rs' --exclude-dir=target --exclude-dir=.git $(consumers)

# Look for parallel emotion types (should use EmotionField/AffectField):
grep -rn 'struct.*Fear\|struct.*Mood\|struct.*Emotion\|fear.*f32' \
    --include='*.rs' --exclude-dir=target --exclude-dir=.git $(consumers)

Audit Step 3 — Classify findings

For each hit, classify:

Classification	Meaning	Action
False positive	The code is legitimately consumer-specific (e.g., MonsterThreatSource impl)	No action — document why
POC simplification	Duplicates substrate but produces identical behavior at POC scale	Document as known debt; fix when scale changes
DRY violation	Re-implements substrate under a different name	File issue; extract to substrate
Architectural violation	Violates two-brain model / sync boundary / domain classification	File issue; redesign

Audit Step 4 — Report

Summarize findings:

## Substrate-first audit (date)
### Substrate inventory
- [domain]: [primitive] at [location]
### Findings
- [file:line] — [classification] — [description]
### Clean
- [domain] — no violations found

---

The vocabulary-translation defense (why this skill exists)

The hardest failures to catch are the ones where the substrate exists but
ships under a name that doesn't match the concept you're searching for. This is
the R296 canonical failure (documented in AGENTS.md §"Manifold Geometry"),
applied internally:

You think: "I need a threat field"
You grep:   "threat field" / "spatial hash"  → 0 hits
Substrate:  GenericSpatialBelief + decay_confidence  → exists, fully functional

You think: "I need emotion decay"
You grep:   "emotion decay" / "fear fade"  → 0 hits
Substrate:  tick_feeling_brain + DecayRates + EmotionBaseline  → exists

You think: "I need state persistence"
You grep:   "save state" / "persist"  → 0 hits
Substrate:  LocalKvStore + freeze_avatar_delta + ShardIndex  → exists

The defense: always search 3+ vocabulary variants. The translation table
in Mode 1 Step 1 is the canonical reference. Extend it when you discover new
mismatches.

---

When NOT to use this skill

• Pure refactoring that doesn't add new concepts (renaming, reorganizing)
• Bug fixes in existing code (the substrate is already consumed or not)
• Test-only code (tests can define inline helpers)
• Build/config changes (Cargo.toml, scripts)

---

Relationship to sibling skills

Skill	What it checks	When
substrate-first (this)	"Does the substrate already exist? Are you duplicating it?"	Before writing code + audit
boundary-guard	"Is this code in the right repo? Is the consumer too fat?"	After writing code + audit
feature-gate-audit	"Do feature-gate claims match source wiring?"	Before promoting/demoting flags
goat-audit	"Has the katgpt-rs primitive been cherry-picked to riir-*?"	Cross-repo cherry-pick tracking
doc-sync	"Do docs match git history?"	After landing plans/issues

substrate-first is upstream of boundary-guard: if substrate-first
catches the drift before it ships, boundary-guard has nothing to find. They're
complementary — substrate-first is the prevention, boundary-guard is the cure.

---

Filing violations

When the audit finds a DRY violation or parallel system:

1. File an issue in the repo where the violation lives
2. Reference this skill + the substrate it duplicates
3. Include the vocabulary translation (what you searched for vs. what the
   substrate is actually called)
4. Propose the fix (consume substrate / extract to substrate / revert)
5. Classify (POC simplification vs. DRY violation vs. architectural violation)

Do NOT fix in the same commit as detection — separate detection from fix so
other agents can review the violation independently.

---

Run log

Compacted 2026-09-11 (user-directed verbose-history pass; the file was 74 KB).

Pruned to the newest 15 rows 2026-10-12 (the 81920B skill-size ceiling; 39 older rows dropped, and two 10-12 rows that had been mis-inserted into the Mode-1 vocabulary table moved back where they belong). git log -p -- .agents/skills/substrate-first/SKILL.md recovers every dropped row verbatim.
Rows carried full verification narratives until this compaction; git log -p -- .agents/skills/substrate-first/SKILL.md recovers any of them verbatim (git log -S '<date>' -- <this file> for one row). New rows append ONE line each.

Standing lessons (distilled from the compacted rows — load-bearing audit classes)

• Body-read before filing a sigmoid/def hit. A local fn sigmoid that DELEGATES to katgpt_core (e.g. a ±50-clamp wrapper) is the SANCTIONED delegation pattern (ndb Issue 611's own fix shape; riir-auth session.rs; chain runtime.rs) — a definition-grep hit is a lead, not a finding. Corollary: katgpt_core::sigmoid is Cephes (~1 ULP), NOT bit-identical to libm-exp forms — delegation is a semantic change where bit-exactness is committed (freeze-commitment paths carry a version-boundary caveat).
• Documented false-positive classes for distance/sqrt greps: squared-distance comparisons (dx*dx+dy*dy <= r*r — the CORRECT no-sqrt idiom; "fixing" it is a pessimization); wire/snapshot [f32;N] DTO operands (no MapPos3D to delegate); grid-space heightfield math; statistical/financial sqrt (std-dev, AMM LP checked_sqrt); disc radial sampling (r*sqrt(u)); screen-space view-plane placement; #[cfg(test)] sites; closed-form test oracles (PSD max-eigenvalue).
• Copy-gate convention: a justified copy needs in-source rationale + a divergence-failing test (chain Issue 139 — a doc-only mirror with only self-consistency tests is NOT enough). Documented copy families: BLAKE3(pubkey)[..16] (5 gated instances — consolidate the 6th onto riir_wallet_signer::player_id_hash); splitmix64 finalizer (~20 substrate-side module-locals, all bit-identical — consolidate at the 4th CONSUMER-side arrival, currently 1); cross-repo message formats single-source by IMPORT (dapps imports chain's canonical signing_message).
• Feature-isolated local math is REQUIRED, not drift — when the feature pulls no katgpt-core (chain forensic f64, congestion inline, the lora bridge keeping katgpt out of the Docker context). Production-dead surface carrying latent duplication under an available dep → delegate-or-prune when consumers materialize (curator_bridge class).
• Translate CRATE names too, not just type/fn names — mmorpg vs riir-games-mmorpg missed a substrate consumer the issue itself cited (Issue 870: the proposed additive StatMods was refuted by the EXISTING clamped StatModifier, Plan 411).
• $(var) executes a command named by the variable — the repo-set helper must be a shell FUNCTION called as $(consumers), or the grep silently scans the whole tree.
• Sanctioned model-consumer shapes (not drift): injected-verifier traits (GmSignatureVerifier pattern); parameterized general form + env-toggle special case (quest_center "one filter home"); feature-neutral placement for gate-union decoupling (HeroStats); Arc<AtomicU8> config seams; two-brain occlusion divergence between substrate and chain verifier (AOI — symmetry pinned on the CHAIN type); plane-split crypto with a hard pub(crate) bar + documented nonce-tradeoff rationale (ndb CLI seal/unseal).

Date	Scope	Verdict	Record
2026-10-26b	Mode 2 mini-audit — the post-422db349e delta (the amo lane's 01:10+07 anchor → ~05:5x; 56 fresh .rs across 7 repos: katgpt-rs 13 [the katgpt-core substrate wave: anchored_reach/metabolic_gate/stale_residual/template_decode/analytic_lattice/functional_substitution/lattice_memory/leakage_probe/hint_regret/spectral_pencil + bench_951 + kv-903], ai 16 [the 1127/1128 arc: anchored_spawns + hero_threat_lane + cognition reorg + karma_engine/p2p_gossip + bench_1125 trio + walk_grid], chain 5 [commitment_cache i166 + dispatch/server + 2 test lanes], infer 6 [BPE tokenizer + G1/G2/G4 gates + kv_evict + spike_census bin], remake 8 [the plan-043 P10 native p2p lanes: seeder/p2p_client/p2p_session + sawtooth/goodput/vessel probes], refine 7 [kernel_opt corpus — fixture class], sdk 1 [scenelab trace_index]; signature greps + body-reads; zero-cargo)	CLEAN — ZERO FILINGS: TEXTBOOK consumes ×3 — 1128 anchored_spawns seats patrol_home through the EXISTING home_patrol/leash_frustration/chasing_hero substrate ("no new system, no thresholds" in-source; the unanchored genesis + quest-ring fallback named as the LIVE 1127 producers), hero_threat_lane delegates game_traits::{sigmoid, approach::*} (the shout falloff rides the route-stamp sigmoid family), walk_grid delegates game_traits::sigmoid + geometric grid sqrts (documented class); SUBSTRATE-SIDE births with lineage headers (the ladder, not duplication): stale_residual kl_logits = stable softmax-KL with the ln(1e-30) floor rationale in-doc, input domain distinct from ugc_schedule's probability-space kl_discrete (not a dup), metabolic_gate delegates crate::sigmoid, lattice_memory SHIPS ITS OWN sigmoid-only grep-gate test (comment-stripped needle over every module file — the law as an in-crate gate), leakage_probe/bench LCGs = seeded fixtures per the gate convention, spike_census = SplitMix64(args.seed) seeded; chain node_hash_from_player = the injective zero-pad EXPANSION consuming player_id_hash (chain→guard identity bridge, never serialized, boundary posture in-doc), dispatch/delegation blake3(pubkey)[..16] sites = #[cfg(test)] fixtures carrying the byte-for-byte mirror comment (the documented copy-gate family); remake p2p lanes = a DISTINCT chunk-distribution concern (hunks/slips/verify over str0m datachannels — not game-state transport; str0m/bao already BOUNDARY-declared seal-view-only behind p2p_vessel, rows 166/167), karma_engine pair-key HashMap = the 1064 ordered-pair O(1) law (not coordinate-keyed), commitment_cache = id-keyed papaya over the existing MatrixAccount map, authority_sim's new std-Mutex handles = copies of the established boot-rig pattern (no papaya in-crate), tokenizer merge_ranks = BPE-internal; zero new_v4 / softmax-impl / belief-struct / coordinate-keyed / Mutex-policy arrivals	katgpt-rs (this commit)
2026-10-26	Mode 2 mini-audit — the post-25b delta (amo idle lane; fresh .rs across 5 repos: dapps 6 [the kat_migrate lane: migrate.rs + migrate_sweep.rs + tests + the worker's migrate_contract/migrate_routes], auth 2 [github_pow.rs + measure example], kat 2 [kat_protocol_migrate + crosswalk test], sdk 1 [aoi_fanout_roster_goat bench]; modified: kat-service worker files, auth account_key/bridge, kat_protocol_freeze, sdk aoi_fanout_curator, reflexer vessel env/lib, katgpt-assign trio; signature greps + body-reads; zero-cargo)	CLEAN — ZERO FILINGS: dapps [..16] ×3 = the documented derivation families with the formula in each doc-comment (acct16 "one derivation, two widths"; reward_key row-key; mint_receipt_id "the same derivation the mint plane pins" — cross-plane single-source by re-derivation, pin named in-source); github_pow = pure fetch/verify module, one blake3::hash over canonical serialization (the digest family, formula in-doc), threshold consts not math; aoi curator/bench + katgpt-assign + reflexer vessel = zero signature hits; zero new_v4/splitmix/softmax/Belief-struct/spatial-HashMap/RNG arrivals; tests+corpus = fixture class	katgpt-rs (this commit)
2026-10-25b	Mode 2 mini-audit — the post-25a delta (the 4090 idle lane's own sync window, 3 fresh .rs + 1 py across 1 repo: reflex src/lanes/elephant.rs NEW [412 lines], src/harness/runner.rs + src/bin/harness.rs mod, scripts/elephant_server.py NEW; the rethink-032-T10.2 mammoth/elephant comparison-lane wave 1e6d4db..f60923a; signature greps + body-reads; zero-cargo, sibling measurement active)	CLEAN — ZERO FILINGS: elephant.rs = pure HTTP wire lane (env consts + post/get + serde parse + determinism_view clock-strip) with ZERO numeric-decision surface — no sigmoid/softmax/exp/hash/seed/distance; the TeacherWireLane impl = textbook glue over the EXISTING lane trait (the mammoth-family shape, spend_env/MAX_CASES_ENV spend-ceiling contract carried); runner.rs softmax ×3 = doc-comments DESCRIBING the elephant/bekko lanes' own readouts (never implemented here), distance_only_* = pre-existing gate-floor test fn names, blake3 mentions = gating doc-comments; harness.rs &digest[..16] = the corpus-row blake3 DISPLAY seal in a log line ("blake3 {}…" ellipsis literal — the display-seal family, not a truncation decision); elephant_server.py = zero signature hits (reference qev serve plumbing); zero new_v4/splitmix/Belief-struct/spatial-HashMap/Mutex/emotion arrivals	katgpt-rs (this commit)
2026-10-25	Mode 2 mini-audit — the post-24c delta (anchor 9d27c7c56 end, 10-24 ~23:1x +07 → 10-25; 2 fresh .rs across 2 repos: ai 1 [riir-gpu bench_1122_site_tax_and_ktile64 — the Issue-1122 bench vehicle, sibling lane, READ-ONLY as-found], remake 1 [canvas_draw_placed — the plan-412 3d-iii placed-rows lane]; mod-set 17 across the same 3 repos: render 4 [scene_set/skinned/zone_scene/assemble], remake 12 [the canvas_draw family + authored_life_view/dressing_rig/head_look/meadow_ground/quest_talk_view/vessel_boot/vessel_scene], ai 1 [riir-gpu lib]; signature greps + body-reads; zero-cargo)	CLEAN — ZERO FILINGS; the wave is a CONSOLIDATION, not duplication: seal-scene scene_set.rs carries the single-home law in its header ("the editor's model viewport lays its scenes out through it today; the RUN viewer will, so both lay the same parts out alike") and is consumed by seal-render lib/assemble/zone_scene AND sge's model_3d_view — the plan-412 single-renderer law as shared-home birth; skinned.rs likewise ("one implementation of skinning… fixed operation order, so posed output cannot shift" — Issue-940-style determinism in-source); remake canvas_draw_placed.rs = TEXTBOOK consume (scene_set::{step_animated, resources_worth_posing, PoseSink} + assemble::scene_skin + LandOn wholesale; the idle "by the map page's own sampler"; the giver's gaze arrives as PlacedGaze from head_look's verdict, never recomputed). REST = documented classes: head_look 1−exp(−k·dt) = view-layer ease (seal-view no-dep class) with the in-source router comment sending the placed turn to the canvas; meadow_ground distance3 [f32;3] + skinned quaternion-normalize/segment sqrts = the geometric class; HashMaps = id/class-keyed fade/joint/land state (zero coordinate-keyed); Mutexes = the app-thread sim-state handles (papaya N/A); skinned [..8] = a test-array slice assert (not a hash truncation); bench_1122 = pure harness, zero RNG/seed/sigmoid surface; zero new_v4/splitmix/softmax/belief-struct arrivals. ALSO REPAIRED (this commit): the run-log table itself — the `	---
2026-10-24c	Mode 2 — the post-24b delta wave (anchor 9d27c7c56, 10-24 16:00+07 → ~23:1x; 147 fresh .rs across 12 repos: sge 41 [the model_facing arc: editor-core probe + gizmo + seeds + store tables + run_reach/run_spawn_models + feat-map fire/lava/waterfall/relief], remake 32 [canvas_draw host/effects/water + vessel_scene/map_stream/chunk_stream + eye_order/load_lint/lod_payload/scheduler gate tests], train 14 [bench-algo reasoner + vq_codebook backprop arms — READ-ONLY, routing-restricted], infer 11 [prefill_cuda_full + gemm mma + laya agent/stability + plan622 g1], refine 11 [kernel_opt corpus entries_gpu_101-119 — fixture class], sdk 8 [riir-vessel cas/chunk_manifest/pack/codec — the ndb-632 wire lane's landed state], render 8 [skill_fx canvas + waterfall/ice + scene zone vocab], katgpt-rs 5 [kv960 T2 alloc landing + its gate tests], ai 4 [bench_1113/656/809 test lanes + gpu lib], ndb 4 [artb + vessel query_hash], reflex 5 [elephant lane + distill + harness], seal-std 4 [seal-fx effect sim/billboard/classify]; signature greps + body-reads; zero-cargo)	CLEAN — ZERO FILINGS; the 24b row's WIP flag RESOLVED: katgpt-rs kv_sensitivity_alloc.rs LANDED as substrate-side COMPOSE — a behavioral-damage weight into the SHIPPED waterfill_bits/BitAllocator allocator ("no new search, no new quantizer: one weight vector"; T1's negative-control caveat in-doc) — the extend-via-input form, not a parallel allocator. TEXTBOOK consumes: sge run_reach.rs delegates the reach verdict to riir_hunt_reach::AimReach wholesale ("the answer is not computed here", :331 sqrt = display distance); the seal-fx single-source-by-leaf law HOLDS through the wave — seal-render bevy_canvas_skill_fx + seal-view {skill_fx,fx_player,buff_fx} all consume seal_fx::{effect,skill_fx,buff_fx,goldens}, zero copy arrivals; sdk riir-vessel cas/chunk_manifest = blake3 content-addressing over the artb wire (the 632 wire lane posture re-verified). REST = documented classes: infer gemm_mma golden-ratio LCGs (8 sites, all inside #[cfg(test)] mod tests @5197) + train train.rs:1185 xorshift seed (also #[cfg(test)]) = the fixture-RNG class; train carry.rs:193 momentum_beta σ(b) = the bench reasoner's model-forward parameterization (riir-bench-algo carries NO katgpt-core dep — only optional katgpt-pruners behind carry_parity); train model.rs softmax_inplace = model-forward; ndb vessel query_hash [..8] = cache-key derivation with perf rationale in-source (~80ns vs ~1100ns hit); reflex distill/harness [..16] = display seals; sge build_lane [..8] = warning-text display; remake gate tests hash[..8] = fixture stuffing; HashMaps = id/texture/DB-row/tile-cache keys (zero coordinate-keyed perception); Mutexes = infer prefill GPU buffer/state mirrors (resource handles, papaya N/A) + laya PACKED_ACT_BITS = the documented parity-seam diagnostic ring (cap 4096, aliasing-hazard story in-doc); zero new_v4 / belief-struct / emotion-struct / non-model softmax arrivals	katgpt-rs (this commit)
2026-10-24b	Mode 2 mini-audit — the post-anchor delta only (10-24 ~09:00 → ~16:10; 75 touched .rs across 9 repos, 12 added: katgpt-rs 2 [kv960 T0a/T0b kv_sensitivity + capture], ai 2 [hero_cast + tests], refine 3 [multivector kill-test + router_per_span_dump], train 1 [the 648 pin test — the morning filing's fix half], remake 4 [eye_order + load_lint + g1 tests]; mod-set: sge 25, remake 8, render 8, ai 8, refine 4, katgpt-rs 4, train 2, seal-std 3, reflex 1; signature greps + body-reads; zero-cargo under 5 live sibling lanes)	CLEAN — ZERO FILINGS; 1 PRIOR MENU DISCHARGED: reflex harness/runner/router.rs:111 — the 10-15 NOTE-LEVEL inline f64 sigmoid now DELEGATES exact_sigmoid_f64 with full provenance in-source (the 014 engine.rs precedent + this audit's 10-15 row + katgpt-rs bbc28a091 cited) + the frozen-legacy-body pin test (sigmoid_delegation_matches_frozen_legacy_body: bit-identical x≥0, ULP-bounded below, [−120,120] sweep, the legacy body frozen in-test) — the copy-gate done end-to-end from note to fix. TEXTBOOK: kv960 kv_sensitivity.rs:738 splitmix = the #[cfg(test)] FIXTURE class (in-source "No external RNG dep; same seed → same stream" + the Sterbenz-exact bimodal fixture); refine multivector.rs = modelless kill-test consuming katgpt_core::simd::maxsim_score (substrate wholesale), its [..16] corpus_digest = the frontier_miner short-digest family; train 648 pin test = the expected delegation-pin landing (same-day filing → same-day fix). REST = documented classes: ai constants.rs configured-point sqrt + knockback falloff (the plan-032 lane), sge feat-map/skinned_model/walk_dots geometric [f32;3]/xz math, remake viewer/lib.rs splitmix doc-comments = the landed Issue-057 consolidation verified holding ("the view layer's ONE splitmix64 draw"), render bevy_canvas/grass view-layer geometry, forward 1/√d + rms = model-forward, reflex router [..16] = the routed-artifact seal display id; remake eye_order/load_lint + ai hero_cast = zero signature arrivals (glue lanes over existing vocabulary); kv960 sibling T2 WIP (kv_sensitivity_alloc.rs untracked) measured as-found, unaudited per detection-only	katgpt-rs (this commit)
2026-10-24	Mode 2 — the 10-22 10:46+07 → 10-24 ~09:00 delta wave (anchor 42712e58c; 200 fresh .rs across 13 repos: remake 65 [the 052 quest-server arc + p2p lanes + canvas_draw host], sge 23 [run_mint + body_measure/footfall/trail lanes + legacy_crit/skill seeds], render 8 [assemble/presence/landfield — post-carve arrivals], train 31 [plans 459-461 + damping_probe + elbo_boundary + hybrid_link + vela — read-only], ai 14 [sst_gate + quest_state/effects + strike_law + 1113/1114 benches], infer 15 [Issue-959 golden_nll + tokengate + plan622 keybias gates], ndb 15 [scope lanes + item_instance — the 052 mint arc], katgpt-rs 9 [sigmoid_budget + softmax_fix + sst_certified + lattice_reopen], seal-std 9 [seal-combat + seal-item + seal-fx move_fx/ui_fx/zone_music], sdk 6 [vessel lod/chunk_manifest + e2e consume_chaos], reflex 2 [clef_conf_probe + depth_probe], rethink 1 [enc_g4_alloc], instinct 2 [tetris_cont]; signature greps + body-reads; zero-cargo under 4 live sibling lanes)	1 FILED: riir-train Issue 648 — damping_probe/mixer.rs:135 pub(crate) fn sigmoid(f32) restates the plain-libm body with katgpt-core REACHABLE (hard dep, engram_privilege — the exact posture that made 636 a filing), zero rationale/pin; f32 flavor so the 636 arm-(a) delegation is NOT zero-behavior (fast_sigmoid = Cephes ~1 ULP) — menu delegate-stated	recorded-refusal + twin-pin, filed detection-only (cc7885d5). NOTE-LEVEL ×3: seal-item Draws splitmix = the seal-fx shape missing its ONE-LINE pointer comment (seal-fx sim.rs carries the copy-gate doc done right — "copied because this crate is a leaf… keep bit-identical" — the twin site needs the same line); ndb item_instance.rs:47/:229 derivation prose omits the INSTANCE_ID_CONTEXT prefix the code applies (seal-item's InstanceId is OPAQUE — "128 bits its minter chose" — so no wire bug; the doc formula is just short one hash input); instinct cont_seed = recipe-published posture (the export bin's piece_seed string IS the cross-medium pin). TEXTBOOK: sigmoid_budget.rs = substrate-side birth delegating fast_sigmoid with measured laws in-doc; golden_nll delegates fixed_log_softmax_target_f64 (959 law in-source); binary_kernel/elbo softmax_row = model-forward/training class; scope_counter Mutex<()> = textbook RMW write-latch (race named + chaos test cited, papaya rule N/A — not a map); hero_live/seal-view Mutexes = option cells + app-thread queues (papaya N/A); ndb scope/item_instance [..16] = lane-id/instance-id derivation family; quest-server nonce [..16] = the request-nonce family; body_measure/gizmo/lod/billboard sqrt = geometric vector math over raw [f32;N]; value_replay_log splitmix = the synthetic FIXTURE class ("a KNOWN answer" fixture, in-doc); softmax_fix/sst_certified = the substrate homes themselves; zero new_v4 / belief-struct / spatial-HashMap / emotion-struct arrivals
2026-10-22	Mode 2 — the 10-21 00:34+07 → 10-22 delta wave (anchor 882e5b3a3; 163 fresh .rs listed / 152 body-present after the corpus-fixture class exclusion, across 10 canonical repos: seal-render 87 [the plan-412 canvas carve], sge 17 [the wave's sge half: hit_sound/skill_book/zone_description lanes], ai 26 [the 639 fog lane + the 1097/1099/1102 combat splits + the composition/* reorg], katgpt-rs 7 [delta_mem/rls.rs Plan-630 exact-RLS rung + katgpt-extract bet_search], refine 11 [corpus fixtures], infer 3 [Issue-1106 stability readout], reflex 2 [modal_floor + startlux], rethink 2 [remote_lane v1 client], seal-std 5 [seal-ability/seal-skill books + seal-fx sim], remake 4 [fishing/job_skills/locomotion_config]; signature greps + body-reads; zero-cargo under the sibling sge merge gate at load ~84 on battery)	**CLEAN — ZERO FILINGS; the headline is the carve: the plan-412 move executed the 231 splitmix-home consolidation CLEANLY across the repo boundary — seal-scene look/seed.rs is the single production home (fire/lava/waterfall consume salted_placement_seed with their documented salts), sge's copies are GONE (bevy_canvas_butterfly deleted; editor-types re-exports seal_scene::{… look …} wholesale so editor_types::look::seed still resolves), and the two sanctioned singleton-locals (editor-core services.rs sample_blake3, editor-data random_equip) carry UPDATED doc-comments naming the new home + the keep-bit-identical copy-gate line — a copy-gate family that survived a cross-repo move intact. TEXTBOOK consumes: the 639 fog lane hero_threat_memory delegates game_traits::spatial::{belief_confidence, MIN_CONFIDENCE} (the one-law citation in-source) and HeroItemMemory stores raw (pos, seen_tick) with read-time substrate decay — zero restated decay formulas; stability.rs = the new calibration-only instrument stating the sequential-f32-fold Issue-940 law in-source + brute-force parity gates (its sqrt = cosine/L2 instrument math, structurally forced: riir-infer-laya carries no katgpt dep); rls.rs = SUBSTRATE-SIDE birth (the exact-RLS rung, Plan 630, naming its δ-rule + ridge_solve ladder siblings in-doc — the ladder, not duplication); modal_floor.rs = the integer-counts census (pool → global → case-order tie law, Issue-940 compliance in-source); lava inline σ + fog 1−exp(−k²) = the documented structurally-forced / pinned-transcription classes moving with their files; cognition_steer dx·dx+dy·dy = delta-LENGTH lookahead normalization (the spin_break no-accessor class); fishing_journal radial sqrt = a view-layer display string; zero new_v4 / belief-struct / spatial-HashMap / softmax / blake3-trunc arrivals	katgpt-rs (this commit)
2026-10-21	Mode 2 — the 10-20 ~04:00 → 10-21 delta wave (117 fresh .rs across 14 repos + ~380 mod-set retouches: katgpt-rs 22/24 [kron_memory substrate birth + the NEW katgpt-extract crate], ai 13/56 [command_lane/command_grammar/fishing/spin_break/survival_stoop/skill_forces/breadcrumb_trail], sge 56/194 [plans 393-405: quest_generate/quest_search/quest_pack/shops/combat-power/level-points/sound/fish/net-selector], seal-std 12 [the seal-quest/shop/swim/stat/formula leaf wave], remake 6/102 [fish_cruise/dev_chat/capture_clock/quest_selection/skill_demo/panel_fallback], refine 5 [corpus fixture + scan_prune], train 2 [bench-algo — routing-restricted, read-only], ndb 1 [riir-rag math seam], dapps/llm 1 each; signature greps + body-reads; zero-cargo)	1 FILED: seal-game-editor Issue 245 — editor-data carries TWO identical pub fn sigmoid(f64) (quest_generate/pick.rs:43 + quest_search/types.rs:289, byte-identical bodies, zero rationale, both pub) while 58b6fe90b already built the shared lexical.rs seam for exactly this module pair — menu = one home in the shared module + re-export + a same-body copy-gate test, zero new edges; intra-crate DRY (the 057/231 genus), structurally forced at repo level (editor admits no katgpt package, Plan 393 T1.2) so NOT delegation. REST CLEAN — textbook single-source-by-leaf: seal-swim SeededLcg/FishCruise/Water/MODEL_FORWARD consumed wholesale by BOTH renderers (seal-view fish_cruise + editor-widgets swim_apply — the duplicated fish-swim CONVERTED to the shared zero-dep leaf this wave, the seal-fx strongest form, first-draw constants pinned in-leaf); ndb riir-rag/math.rs = the copy-gate done RIGHT (the katgpt_cmp OFF-arm transcription + a ~4.2M-point bit-identity mirror test compiled only when both arms exist, the embedder's model-id promise cited in-source); ai spin_break = MapPos3D::distance_2d_to consumer (its unit() helper outputs raw [f32;2] bearings — no accessor exists, documented class); survival_stoop sqrt = test oracle; sge llm.rs [..16] = the echo-token display family; school.rs HashMap<(i32,i32),Vec<usize>> = one-shot scatter-time bucket grid (build-time pool ordering, not per-tick perception); katgpt-extract temper.rs sigmoid = RECORDED-REFUSAL rationale in-source (µs-path not ns-path, platform-portable plain exp) + span_pointer.rs module-local = substrate-side same-crate, NOTE-LEVEL one-line menu (cite temper's rationale or delegate); kron_memory LCGs = the unseeded-global-RNG gate convention; train carry.rs softmax = model-forward, Box-Muller = the in-crate RNG class. MOD-SET: editor-core effect.rs:146 VERIFIES the 231 menu LANDED (delegates editor_types::look::seed::splitmix64_finalize); dev_vibe:485/locomotion_gait:66 = the documented seal-view no-dep class; hero_routine det_trig delegates (1057 holding); hero_progress_drive consumes the designated luck.rs home (1044 holding)	sge 245 filing + katgpt-rs (this commit)
2026-10-19	Mode 2 — the post-18d delta wave (anchor 29f37db2c, 10-18 11:28+07; 30 fresh .rs across 6 repos: seal-std 12 [the NEW seal-fx leaf, seal-std issue 001 — schema/curve/classify/quads/goldens], sge 7 [fx_export writer + warp_lint tool + skill_projectiles + look/motion + canvas motion tests + 2 examples], remake 5 [the fx reader arc: fx_player/skill_fx/clip_fx/buff_fx/audio], katgpt-rs 2 [fdr_credit Terminal-RSI + spr_order], ai 3 [buffs.rs plan-033 slice B + pin_gate_tests + wkg_inspect example], dapps 3 [sponsor + stake ingest/routes]; signature greps + body-reads; zero-cargo under 2 live sibling suites)	CLEAN — zero filings; the wave's centerpiece is the STRONGEST single-source-by-IMPORT form on record: seal-fx = canonical leaf (curve.rs header carries the port rationale + the bit-identical-both-players premise), editor-core::effect + editor-release::fx_export + seal-view {fx_player,skill_fx,clip_fx,buff_fx} all RE-EXPORT/import seal_fx types (editor-widgets consumes transitively via editor_core — no leaf dep needed), goldens (SKILL_FX_WIRE/CLIP_FX_WIRE/BUFF_FX_WIRE) pinned IN the leaf and parsed by BOTH halves' tests — no copy exists to diverge, stronger than a divergence pin (the 232/233 class PREVENTED, not cured). TEXTBOOK consumes: buffs.rs = swarm::luck::seeded_hash designated home + substrate Hp/Position/is_dead + determinism laws in-source (the 1044 fix holding through its first consumer wave); pin_gate_tests = MapPos distance_2d_to accessor; NEW substrate-side operators, not duplicates: katgpt-pruners mm_squash = Michaelis-Menten reciprocal form (P/(P+D') — saturating, NOT logistic; no katgpt-core restatement found, substrate-home birth, house sigmoid-rule compatibility cited), fdr_credit.rs cites the percentile tail-support floor in-source (audit culture compounding); dapps sponsor.rs [..16] = the 81-byte wire codec AccountId family; audio.rs linear-over-sigmoid = in-source rationale (two-anchor contract). Zero splitmix/new_v4/belief-struct/spatial-HashMap/emotion arrivals; editor-core's 1-line lerp = documented cross-reference to the seal_fx private copy, below the gate	katgpt-rs (this commit)
2026-10-18d	Mode 2 — the post-18c delta wave (41 unaudited fresh .rs across 8 repos + 2 excluded-by-class: jup-tool 2 [the 4cf74d2 attribute-only posture], ai 10 [the 1091 spawn-HP + pvp-team tails], dao 2 [the landed Proposal-009 orphan], dapps 8 [the Proposal-009 TUNA airdrop trio + explorer views], mmorpg 2 [the 112/113 fixes], train 10 [plan410 stage0 + discoloop POC — READ-ONLY, routing-restricted], remake 3 + sge 3 [the plan-390 molten twin wave M9/M11 + the reach-cap fix]; katgpt-rs 223 = the 947 comment-move docs commits [zero semantic arrivals — removed non-comment lines = 3 markdown lines], riir-kat 35 = a whole-crate rustfmt round-trip [byte-identical, fixture class]; signature greps + body-reads; zero-cargo under sibling builds)	1 FILED: seal-game-editor Issue 233 — the vessel-lava wire literal pin is ONE-SIDED: the viewer's the_wire_lava_reads_as_the_editor_pressed_it freezes the full WIRE JSON (uv_per_tile 0.5, warm_share 0.35, embers 36, limelight 8/20/4, glow cap 10.0) but the EDITOR side asserts no literal (its 7 tests = round-trips + const-vs-const) — an editor-types retune reds nothing anywhere while silently staling the viewer's frozen press; the plan-300 both-repos law half-satisfied; the chain-139 class third instance in this lane family (232 flow-turn bodies, remake 102 fade constants, 233 wire literals); menu = the editor twin test. VERIFIED-STILL-VALID: remake Issue 102 unchanged by the wave's re-touch (viewer FADE_GAIN 8.0 / FADE_STEPS 32.0 no-snap vs the editor's 12/16 end-snapped — the filing's table stands). REST CLEAN — textbook consumes: hero_progress_drive = game_traits::sigmoid delegate ×3 + the luck::{seeded_hash, splitmix64_finalize, SPLITMIX64_GAMMA} designated home (the 1044 fix holding); distill_tsd:1094 = a closed-form test oracle (test_row_confidence_matches_commit_arithmetic); discoloop Box-Muller/RMSNorm/Adam/1-√n sqrt = model-forward + POC in-crate RNG with in-source rationale; meadow_lava:283 inline σ = structurally forced (no katgpt dep, the documented seal-view class) + the 102-covered surface; the plan-390 reach cap = the CORRECT twin shape (reach_cap twin, asserted both sides); meadow_lava HashMap<(LavaPart,u32)> = a render mesh cache; dapps [..16] = the doc-comment display family; belief/emotion/new_v4 = zero arrivals	sge 67727d54 filing + katgpt-rs (this commit)
2026-10-20a	Mode 2 wave audit (10-18 12:00 → 10-20 ~04:0x UTC; union-of-touched .rs via git log across the 9 workspace repos: katgpt-rs 18 / ai 1444 / train 711 / refine 55 / infer 125 / instinct 2 = ~2355 files, the bulk = the 2026-10-19 07:58 FLEET FMT SWEEP + the ai-1095 soak lane + refine B259-B264 corpus; signature greps softmax / new_v4() / fn sigmoid / SpatialBelief-family / coordinate-keyed HashMap over the touched sets, body-reads on every hit; zero-cargo, StartLux resident)	CLEAN — ZERO FILINGS: new_v4 hits ×3 = doc-comments CITING the law (pet_roster / motivation types / entity types); engine sigmoid defs ×3 (limelight:111 / arbitration:433 / thermal_lod:408) = the ai-1069 dispositioned NaN-guard + fast_sigmoid DELEGATION wrapper shape verbatim; katgpt-core lib.rs = the substrate home itself; infer elementwise sigmoid_f32 + clef softmax mentions = fmt-sweep-touched pre-existing + doc-comments describing the REFERENCE model's internals; SpatialBelief hits = the known ai-1051 census entry + TEXTBOOK GenericSpatialBelief consumes (flee_steering/emotion); HashMap-spatial hits = id→pos value maps (u32 keys), not coordinate-keyed; corpus entries = fixture text. NOTABLE WAVE CONTENT (no drift): the fmt sweep normalized the fleet (line-ending + style only, blame-verified on the new_v4 sites), B261-B264 landed kernel_opt corpus, ai-1095/plan-638 lanes	katgpt-rs (this commit)
2026-10-18c	Mode 2 mini-audit — the post-18b delta (10-17 21:00Z → 10-18 09:3x: 9 fresh .rs across 4 repos: ai 3 [crowd_state_windows T1–T4 files], refine 2 [B249 slice2/3 corpus fixture class], sge 2 [plan-390 M3b/M4 canvas lava-flow + molten-rig], remake 2 [M9 RUN viewer twins meadow_lava_flow + molten_rig]; signature greps + body-reads; zero-cargo under the plan-436 GPU chain)	1 FILED: seal-game-editor Issue 232 — the plan-390 M3b lava flow-turn landed as a CROSS-REPO TWIN PAIR with zero divergence pins: editor RIIR_DRIFT/RIIR_LAVA_FRAGMENT/flow_turn body/26-line FLOW_LAVA_VERTEX_WGSL vs viewer meadow_lava_flow.rs + shaders/flow_lava.wgsl verified BODY-IDENTICAL this audit, but each side's tests assert only its own body (self-consistency) — the chain-139 copy-gate law (acknowledgment present, pin absent; the seal-remake 102 shape one layer up, pre-drift). Structurally forced (seal-view has no editor-widgets dep — separate deploy surfaces), so menu = per-repo byte-literal pin tests, NOT delegation. REST CLEAN: state_windows.rs = TEXTBOOK consume re-verified on the full body (katgpt_core::event_state_window + habituation_filter::settling_ticks wholesale; det_trig::ln64/exp delegates; EmotionField read-only accessor; zero alloc scratch contracts in-source; the 18b row's verdict stands); molten_rig twins (editor canvas vs viewer rig) = the WGSL-is-the-source design with the viewer's literal-quoted pinned contract tests + deliberate rig-limit divergence — sanctioned, not the 232 class; quantise ×3 private module-locals (canvas fire/molten + remake meadow_fire) = pre-231-census family, below every gate (single-copy-per-repo, distinct steps args); B249 corpus = fixture class	sge 232 filing + katgpt-rs (this commit)
2026-10-18b	Mode 2 mini-audit — the post-row delta only (since 21:00Z: ai 4 [crowd_state_windows T1–T4 landing + G4 gate fix — self-audited at landing via the ai-1081 substrate-check block], sge 11 [plan-356 B2/B3/B4 motion lanes: look/fire	lava	waterfall
2026-10-18	Mode 2 — the 10-16 08:00 → 10-18 delta wave (128 fresh .rs across 13 repos: train 18 [vq_codebook arc + plan445 benches + drive_constants_train], editor 50 [lava/relief/sound/skill-script lanes], remake 13 [meadow_lava/relief/warp/spawn_focus], ai 12 [hero_no_gain + steer_avoid + spawn_place/target_assign + cluster_rebalance + wanderlust tests], infer 9 [riir-infer-clef lane], katgpt-rs 9 [qrr arc + approx + assign explain], dapps 3 + kat 2 + refine 2 [tuna airdrop trio], instinct 4 [eq9], shader 4 [lava gamefx], reflex 1 [gate_drift], seal-std 1; signature greps + body-reads; zero-cargo under Plan-337)	2 FILED: riir-train Issue 636 — drive_constants_train.rs:199 sigmoid64 restates katgpt_core::exact_sigmoid_f64 BIT-IDENTICALLY (same two-branch stable form, binding name only differs) with the home REACHABLE (katgpt-core = hard workspace dep, engram_privilege) + house-law comment present but NO delegation refusal named + NO divergence pin = the 163 shape (copy-gate, NOT the 002 numerics class — delegation zero-behavior, verdicts replay through hard_surface); menu delegate	recorded-refusal + twin-pin. seal-remake Issue 102 — the lava fade twins carry NO literal pin: editor ring_fade σ(12·(edge−d)/width)/steps16/end-snapped vs meadow_lava part_fade σ(8·(edge−d)/width)×presence/steps32/no-snap/presence-pre-quantise — same structural design, DRIFTED constants, WYSIWYG divergence in a lane whose premise is "the way the canvas draws it"; the header's byte-shared-shader claim covers WGSL only, the CPU fade is a duplicated transcription the pin law (plan 300) demands pins for; menu align

References

• AGENTS.md §"Spatial Cognition (Two-Brain Model)" — the canonical perception rules
• AGENTS.md §"Manifold Geometry (Stokes Calculus)" — the R296 vocabulary-translation
  failure pattern (this skill extends it from paper→code to codebase→codebase)
• AGENTS.md §"Latent vs Raw Space Rules" — domain classification
• riir-games-shared/src/game_traits/spatial.rs — GenericSpatialBelief<T> substrate
• riir-games-shared/src/game_traits/ — vocabulary translation table source
• Issue 047 (riir-mmorpg-examples) — the ThreatField canonical failure
• Issue 490 + Issue 493 (riir-game-sdk) — the orchard/motivation canonical failure
• boundary-guard skill — sibling (boundary enforcement, post-hoc)