🧫 Lab

Live validation results, spring science hubs, and sovereign compute access. 13 primals, 13,648+ checks, 8 springs, full provenance. Run it yourself or review the evidence.

The lab is where ecoPrimals science gets validated on real hardware. Everything here ran on a sovereign compute node (i9-14900K, 96 GB DDR5, RTX 4070 / RTX 3090 / Akida NPU) through a live 13-primal NUCLEUS composition. Every result carries a cryptographic provenance chain: BLAKE3 content hashes → rhizoCrypt DAG → loamSpine ledger → sweetGrass ed25519-witnessed braid.

Security status: All 13 primals default 127.0.0.1 bind. BTSP Phase 3 AEAD on all connections. NestGate method-level auth gating. JH-0 MethodGate adopted 13/13 primals. 445 registered capability methods. Zero DEBT markers (primalSpring Wave 35).

Spring Science

Each spring validates a scientific domain. These pages tell the full story — what was reproduced, what was discovered, and what it proved about the infrastructure.

Spring	Domain	Checks	Papers
wetSpring	Life science, metagenomics, PFAS	5,707+	63/63
hotSpring	Plasma physics, lattice QCD, spectral	697+	10+
airSpring	Precision agriculture, irrigation	2,777+	57
healthSpring	PK/PD, microbiome, biosignal, drugs	795	7 tracks
groundSpring	Measurement noise, uncertainty, Anderson	965+	29 baselines
primalSpring	Composition parity, deploy graphs, NUCLEUS	666	— (meta-spring)

2 more springs (neuralSpring, ludoSpring) are documented in the Spring Catalog and are being expanded by upstream contributors.

Total across all 8 springs: 14,314+ quantitative checks, 70+ peer-reviewed papers reproduced, 15 researchers across 9 departments.

Validation & Provenance

Validation results — 235+ structured science checks across 8 workloads, dispatched through ToadStool on a live composition. Real NCBI data (11.9M paired-end reads) processed through both Python and Rust pipelines. Python→Rust parity at machine-epsilon precision.

Provenance evidence — every artifact content-addressed, every pipeline step tracked in a DAG session, committed to a permanent ledger, and witnessed with ed25519 signatures. The braid is PROV-O compliant with DID attribution.

Reproduce it yourself — step-by-step instructions to stand up the same composition on your own hardware and run the same workloads. No cloud. No institutional access. Commodity hardware.

Reproduce Results — step-by-step guide
Provenance Pipeline — how results are tracked and verified

The Validation Pattern

Published results (papers, databases, NCBI)
        ↓
Python / established tools (QIIME2, SciPy, R vegan/phyloseq)
        ↓
Rust implementation (wetSpring, barraCuda)
        ↓
NUCLEUS composition dispatch (toadStool execute)
        ↓
Provenance chain (BLAKE3 → DAG → ledger → braid)
        ↓
Parity check + gap report

Each arrow is independently verifiable. The Rust matches the Python. The composition matches standalone Rust. Gaps are documented and flow upstream. Every successful workload is proof that the deploy graphs, BTSP encryption, discovery hierarchy, and provenance pipeline work in production.

Public Notebooks

Interactive Jupyter notebooks that visualize baseCamp science. Each notebook loads frozen experiment data (JSON artifacts) from the spring repositories — no live primals required.

Notebook	Spring	Story
16S Pipeline Validation	wetSpring	Flagship 16S pipeline, Galaxy/QIIME2 parity, R/vegan cross-validation
Python vs Rust vs GPU	wetSpring	Benchmark evidence: timing, energy, speedup across three tiers
63/63 Paper Reproductions	wetSpring	5 researchers, 6 tracks, full evidence map
Cross-Spring Connections	wetSpring	79 barraCuda primitives, constraint-driven discoveries
Soil Anderson Deep Dive	wetSpring	Anderson localization in soil biology — physics meets ecology
Composition Validation	groundSpring	6 deploy graphs, guideStone Level 3, verb reconciliation
Benchmark Comparison	groundSpring	Rust vs Python timing, three-mode GPU, 110 barraCuda delegations
Ecosystem Evidence	groundSpring	35 experiments, 10 domains, gap resolution, security posture
Cross-Spring Connections	groundSpring	5 primals consumed, 7 ecosystem flows, patterns pioneered
Measurement Science Deep Dive	groundSpring	Five pillars, 13-tier tolerance architecture, Anderson thread
Composition Validation	primalSpring	Deploy graphs, bond types, profiles, discovery tiers
Benchmark Comparison	primalSpring	Rust vs Python timing, energy, guidestone phases
Ecosystem Evidence	primalSpring	85 experiments, gap resolution, security timeline
Cross-Spring Connections	primalSpring	Primal consumption matrix, ecosystem flows
BTSP Security Deep Dive	primalSpring	Per-primal posture, convergence arc, discovery hierarchy

Run them yourself: Clone the spring, cd notebooks/, jupyter lab. Or access via JupyterHub.

Compute Access

JupyterHub provides multi-user notebook access to the live 13-primal composition. Every notebook runs against real primals, not mocks. Three access tiers (observer, reviewer, user) ensure open science with appropriate boundaries.

Compute Access — tiers, architecture, how to connect

For PIs and Reviewers

The lab is the evidence record for the projectFOUNDATION protocol. If you’re evaluating ecoPrimals for institutional adoption:

Review the spring science hubs — each spring page shows what was reproduced and how
Check the provenance pipeline — every result is content-addressed with cryptographic chains
Request reviewer access — read-only JupyterHub access to see the live workspace
Run it yourself — the reproduction guide works on commodity hardware, no institutional access needed

The shared workspace at /shared/abg/showcase/ contains polished results ready for institutional review.

For ABG Members

If you’re in the Accelerated Bioinformatics Group, the lab is also your on-ramp. The same pipelines that produced these results are available through JupyterHub via the compute sharing tunnel. Your workloads run on the same composition, with the same provenance. Your science validates the infrastructure; the infrastructure validates your science.

See Compute Access for how to connect, or Reproduce It Yourself to run this on your own hardware.