neuralSpring Validation Summary

ML primitives and sovereign structure prediction — 4,900+ checks, Isomorphism Theorem, all 6 AlphaFold primitives validated

neuralSpring — sporePrint Validation Summary

Session: S225 | Date: Jun 6, 2026 | Version: 0.1.0 | Handoff: V181 Gate: southGate | Live validation: 9/13 primals via UDS Tier: 2 (sporePrint: frozen data + notebooks + paper baselines)

Headline Numbers

Metric	Value
Workspace tests (IPC-first)	754
Proptest properties	24
Python baselines	397/397 PASS
Rust+GPU checks	4,500+
Total validation checks	4,900+
Binaries	269 (244 validate, 18 bench, 7 other)
Experiments	134 across 11 domains
Papers reproduced	27 (6 faculties)
Capabilities	45 (12 domains)
Named tolerances	233
guideStone	30/37 PASS, 6 SKIP (live southGate deployment)
BTSP	13/13 mandatory
PRIMAL_GAPS	29 main (29 resolved)

Code Quality

Check	Status
`cargo clippy --workspace` (pedantic+nursery)	0 warnings
`cargo fmt --check`	0 diffs
`cargo doc --workspace --no-deps`	0 warnings
`cargo deny check`	clean
`#![forbid(unsafe_code)]`	workspace-wide
`#[allow()]` attributes	0
TODO/FIXME/HACK	0
Mocks in production	0

Performance

Metric	Value
Rust vs Python geomean	38.6x (15 domains)
Fastest speedup	1,104x (multi-objective)
CPU-Python parity	41/41 PASS (1e-10)
GPU max speedup	104x (transformer medium)
GPU coverage	~97%
Multi-GPU parity	384/384 bit-identical
Dispatch overhead	<=1.04x (9/10 ops)

Key Validation Binaries

validate_isomorphism — 6-primitive decomposition
validate_gemm_attention — core neural primitives
validate_dispatch_parity — multi-GPU bit-identical
validate_helixvision — Evoformer, IPA, diffusion
validate_all — full validation suite (244 binaries)
neuralspring_guidestone — guideStone Level 5 (19 certification tests)
validate_ltee_b3_allele_trajectory — LSTM+HMM+ESN allele classifier (16/16)
validate_ltee_b4_citrate_esn — ESN citrate early-warning (16/16)

Notebooks — sporePrint

#	Notebook	Focus
01	`01-composition-validation.ipynb`	Deploy graphs, bond types, capabilities, discovery tiers
02	`02-benchmark-comparison.ipynb`	Rust vs Python timing, GPU speedups, guideStone phases
03	`03-ecosystem-evidence.ipynb`	134 experiments, gap resolution, security posture
04	`04-cross-spring-connections.ipynb`	Primal consumption matrix, ecosystem flows
05	`05-btsp-security-deep-dive.ipynb`	Per-primal BTSP posture, security convergence arc

Notebooks — Paper Baselines (2 faculties, 8 notebooks)

Publishable-grade Jupyter notebooks with full inline Python/NumPy implementations of peer-reviewed science. Each notebook is the math validation base — the foundation layer that Rust, GPU, and primal IPC are validated against. Self-contained; executable on JupyterHub without the neuralSpring repo.

Batch 1: Dolson Faculty (Evolutionary Computation)

Paper	Notebook	Citation	Checks
011	`paper-011-counterdiabatic-evolution.ipynb`	Iram, Dolson et al. (2020) Nature Physics 17:135-142	11/11
012	`paper-012-modes-toolbox.ipynb`	Dolson et al. (2019) Artificial Life 25(1):50-73	9/9
013	`paper-013-eco-dynamics.ipynb`	Dolson & Ofria (2018) GECCO ’18 Companion	7/7
014	`paper-014-directed-evolution.ipynb`	Dolson, Banzhaf, Ofria (2022) eLife 11:e79665	8/8
015	`paper-015-swarm-robotics.ipynb`	Foreback, Bohm, Dolson (2025) IEEE	11/11

Batch 2: Liu Faculty (HMM & Phylogenetic Inference)

#	Notebook	Paper	Checks
016	`paper-016-hmm-phylo.ipynb`	Liu et al. (2014) PLoS Comp Bio	10/10
017	`paper-017-sate-alignment.ipynb`	Liu et al. (2009) Science	8/8
018	`paper-018-introgression.ipynb`	Liu et al. (2015) PNAS	8/8

Total: 8 notebooks, 72/72 checks PASS, 3,337 lines of validated Python source. Faculties: Emily Dolson (Evolutionary Computation), Kevin Liu (Phylogenetic Inference). Remaining batches: 19 papers across 4 additional faculties.

Frozen Data

File	Contents
`validation-state.json`	Test counts, capabilities, code quality, guideStone
`experiment-catalog.json`	134 experiments, 6 faculties, validation tiers
`security-posture.json`	BTSP, cargo-deny, unsafe, BLAKE3 checksums
`cross-spring-matrix.json`	8 primal dependencies, proto-nucleate
`benchmark-data.json`	Rust vs Python, GPU, multi-GPU, isomorphic primitives
`gap-status.json`	28 gaps, 28 resolved
`paper-baselines.json`	8 paper notebooks, 72 checks, 2 faculties, BarraCUDA mappings

Ecosystem

Edition: 2024
MSRV: 1.87
barraCuda: v0.4.0
primalSpring: v0.9.27+ (Wave 46, 458 methods)
genomeBin: v5.1 (46 binaries, 6 target triples)
Bond type: Metallic
Trust model: InternalNucleus
Proto-nucleate: 7 validation capabilities, 6 primal dependencies
Isomorphism Theorem: all neural architectures decompose into 6 primitives (GEMM, Attention, Normalization, Nonlinearity, Reduction, Gating)
helixVision: sovereign AlphaFold2/3 structure prediction primitives