pseudoSpore — Sovereign Science Data

519 GB of real science data with full cryptographic provenance. Downloadable and verifiable on commodity hardware. No cloud. No trust required.

Real science data. Sovereign hardware. Full provenance. Verify it yourself.

westGate has ingested 519 GB across 17+ scientific domains through the complete CAS + Provenance Trio pipeline. Every object is content-addressed (BLAKE3), DAG-tracked (rhizoCrypt), ledger-committed (loamSpine), Ed25519-signed (bearDog), and attribution-braided (sweetGrass). 100% provenance coverage.

This is not a plan. This data is on disk, hashed, signed, and queryable right now on a Ryzen 9 / 96 GB DDR5 / 50.7 TB ZFS raidz1 machine running in a basement.

Data Catalog

Dataset	Size	Objects	Domain	Springs	Provenance
LINCS L1000 Level 5	19.86 GB	6	Gene expression (473K signatures x 12K genes)	wetSpring, tideGlass	5/5 FULL
ChEMBL 37	~15 GB	2	Drug discovery (2.9M compounds, 24.5M bioactivities)	healthSpring, tideGlass	5/5 FULL
PDB	361 MB+	506+	Structural biology (protein structures)	hotSpring, neuralSpring	5/5 FULL
UniProt Swiss-Prot	764 MB	3	Protein sequences	wetSpring, hotSpring	5/5 FULL
ZINC20 SMILES	160 MB	110	Compound screening	healthSpring, tideGlass	5/5 FULL
GTEx V8	2.4 GB	4	Tissue expression	wetSpring, healthSpring	5/5 FULL
SILVA 138.1	188 MB	1	16S taxonomy reference	wetSpring	5/5 FULL
PhysioNet MIT-BIH	22 MB	1	Biosignals (ECG)	healthSpring	5/5 FULL
MassBank NIST	63 MB	1	Mass spectrometry reference	wetSpring	5/5 FULL
NOAA GHCND	11 MB	2	Weather station data	groundSpring, airSpring	5/5 FULL
LTEE REL606	5.8 MB	1	Microbial evolution reference	wetSpring	5/5 FULL
Total (this table)	~38.2 GB	4,752	6 domains	8 springs	100%

This table shows the original pseudoSpore data catalog. The full Data Braids catalog now covers 519 GB across 130+ datasets and 17+ domains.

What “5/5 FULL” Means

Every data object passes through five provenance stages. Each is independently verifiable:

Stage	Primal	What It Proves
1. Content hash	nestGate	BLAKE3 hash — the object IS what it claims to be
2. CAS storage	nestGate	Content-addressed — identity is the hash, not a filename
3. DAG tracking	rhizoCrypt	Parent/child relationships — where data came from
4. Ledger commit	loamSpine	Immutable record — the object existed at this time
5. Ed25519 witness	sweetGrass	Cryptographic signature — who committed it and when

The chain is end-to-end. No stage trusts the previous one — each is independently checkable.

What a pseudoSpore Contains

A pseudoSpore is a downloadable archive carrying data + its proof chain:

pseudospore-chembl37/
├── data/                      # The science data
│   ├── chembl_37.sdf.gz
│   └── chembl_37_sqlite.tar.gz
├── provenance/                # The proof chain
│   ├── blake3_checksums.txt   # BLAKE3 hashes of every file
│   ├── cas_manifest.json      # nestGate CAS object IDs
│   ├── dag_proof.json         # rhizoCrypt DAG chain
│   ├── spine_entry.json       # loamSpine ledger entry
│   ├── ed25519_signature.json # bearDog signature
│   └── attribution_braid.json # sweetGrass W3C PROV-O attribution
├── validate.sh                # Run this to verify everything
└── README.md                  # What this is, where it came from

./validate.sh checks every hash, every signature, every chain link. You are the verifier. No trust required. No central authority. The provenance travels with the data.

Verify It Yourself

Three levels of verification, zero trust required:

Level 1: Check a hash

# Download any dataset file
# Run b3sum (BLAKE3 CLI) and compare against the manifest
b3sum chembl_37.sdf.gz
# Compare output against blake3_checksums.txt

Level 2: Verify the signature chain

# Inside a pseudoSpore archive:
./validate.sh
# Checks: BLAKE3 hashes → CAS IDs → DAG parents → spine entry → Ed25519 sig
# Output: PASS or FAIL for each stage

Level 3: Reproduce on your own hardware

Deploy NUCLEUS on your machine (see Getting Started), ingest the same public datasets, and compare your provenance chain against ours. The hashes will match. The science is deterministic. The hardware is commodity.

The Hardware

All of this runs on:

westGate: i9-14900K, 96 GB DDR5, 50.7 TB ZFS raidz1
Cost: ~$6K hardware + $485/month operating (electricity, ISP, VPS)
OS: Linux (NixOS)
Network: Sovereign mesh via Tower Atomic (bearDog + songBird + skunkBat)

No cloud. No AWS. No institutional compute allocation. Commodity hardware you could build from Micro Center parts.

Spore Taxonomy

Class	Self-sufficient?	What it carries
coldSpore	No	Static marker + frozen data snapshot
liveSpore	Partial	+ Journal + refresh mechanism
pseudoSpore	No	+ Provenance braids, receipts, derivation configs — proves the mountain was climbed
lithoSpore	Yes	+ Runtime + binaries + full data — carries everything to reproduce independently

The spore can’t carry the mountain, but it proves the mountain was climbed.

pseudoSpore #2: Computed Science (hotSpring SU(2) Gauge Theory)

The data catalog above shows ingested reference data — public datasets pulled through the CAS + Provenance pipeline. But the system doesn’t just store science. It produces science.

hotSpring SU(2) — Lattice Gauge Theory is the second pseudoSpore: original SU(2) lattice gauge theory trajectories computed on strandGate using both NVIDIA RTX 3090 and AMD RX 6950 XT. Multi-vendor, DF64 precision, full provenance. The same WGSL shaders run on both GPUs — cross-GPU plaquette agreement within 3.1×10⁻⁹.

arXiv preprint: under refinement. Validation experiment queue in progress. Rung 1 of 6 toward vendor-agnostic lattice QCD (ORCID 0009-0004-2141-0321).