GEMM consolidation v1.5 — Wave 0 measured: 0k passes, 0j kills the density framing, 0l falsifies prefix tracts on physical addresses - #304

Merged

AdaWorldAPI merged 16 commits into

masterfrom

claude/great-curie-d2ufyl

Sep 6, 2026

Commits on Sep 5, 2026

plans: gemm consolidation v1.1 — Mississippi Queen amendment + Wave 0 static results
claude
committed
6f489ec
plans: gemm consolidation v1.2 — grey/white matter over one packed register; third D-GTM-5 correction
claude
committed
dc89eca
plans: gemm consolidation v1.3 — substrate == mask geometry == projection surface
claude
committed
8c9e80f
probe: W0 D-GTM-0g/0h/0i/0j/0k — mask/trie vs GEMM measured; 0j falsifies the density framing
claude
committed
284b999
probe: fix clippy needless_range_loop; blackboard entry for the W0 measured results
claude
committed
701314a
probe: D-GTM-0l — packed-prefix tract coverage measured on the R2IL 6502 ore
claude
committed
77718bc
probe: D-GTM-0n — ternlog amortization, residency knee, and the mask/sparse density crossover
claude
committed
741e34b
probe: D-GTM-0m — behavioural soak across ISAs, and the null that deflates its headline
claude
committed
f1f4023
probe: drop two dead locals left by deduplicating the alphabet control
claude
committed
ab3359f
probe: remove the duplicate alphabet binding
claude
committed
b861892
probe: drop the duplicated alphabet-saturation block
claude
committed
ed75c44
plan: plastic-activation-seam research map v1 (sections B-L; A pending the physics audit)
claude
committed
e8927e1
plan: add the direction control — cross-ISA transfer is asymmetric
claude
committed
236b849
plan: section A — shipped activation physics, from source
claude
committed
babf19d
fix: gate the ternlog probe behind the std feature
claude
committed
a1880fa

Commits on Sep 6, 2026

review: address CodeRabbit findings on the Wave-0 probes and plan
claude
committed
3e598a1

Navigation Menu

GEMM consolidation v1.5 — Wave 0 measured: 0k passes, 0j kills the density framing, 0l falsifies prefix tracts on physical addresses - #304

GEMM consolidation v1.5 — Wave 0 measured: 0k passes, 0j kills the density framing, 0l falsifies prefix tracts on physical addresses#304
AdaWorldAPI merged 16 commits into
masterfrom
claude/great-curie-d2ufyl

Commits on Sep 5, 2026

plans: gemm consolidation v1.1 — Mississippi Queen amendment + Wave 0 static results

plans: gemm consolidation v1.2 — grey/white matter over one packed register; third D-GTM-5 correction

plans: gemm consolidation v1.3 — substrate == mask geometry == projection surface

probe: W0 D-GTM-0g/0h/0i/0j/0k — mask/trie vs GEMM measured; 0j falsifies the density framing

probe: fix clippy needless_range_loop; blackboard entry for the W0 measured results

probe: D-GTM-0l — packed-prefix tract coverage measured on the R2IL 6502 ore

probe: D-GTM-0n — ternlog amortization, residency knee, and the mask/sparse density crossover

probe: D-GTM-0m — behavioural soak across ISAs, and the null that deflates its headline

probe: drop two dead locals left by deduplicating the alphabet control

probe: remove the duplicate alphabet binding

probe: drop the duplicated alphabet-saturation block

plan: plastic-activation-seam research map v1 (sections B-L; A pending the physics audit)

plan: add the direction control — cross-ISA transfer is asymmetric

plan: section A — shipped activation physics, from source

fix: gate the ternlog probe behind the std feature

Commits on Sep 6, 2026

review: address CodeRabbit findings on the Wave-0 probes and plan