01
Source reconstruction
Chain data is read directly from node RPC and normalized while preserving native object structure. Nothing is flattened into a generic transaction abstraction.
Company
Measurement before interpretation. This is the full path from raw chain data to a finding a client can rely on.
Pipeline
Stages
01
Chain data is read directly from node RPC and normalized while preserving native object structure. Nothing is flattened into a generic transaction abstraction.
02
Objects are described by purely mathematical features under a versioned schema. Identical input ranges always produce identical feature output.
03
Relationships come from actual input/output consumption evidence, with explicit boundary semantics at the edges of the indexed range.
04
Populations are summarized with robust methods so any single observation can be positioned against its cohort rather than judged in isolation.
05
Runs are re-verified against digests and manifests. Integrity failures halt the pipeline instead of being smoothed over.
06
Every published number resolves to a dataset, run identifier, block range and verification record.
Live
Live values read from the research backend. Nothing in this panel is simulated.
Reading pipeline state…
Technical detail lives in the research platform — Scale Lab, Verification, Diagnostics.
Evidence first
Every claim on this site maps to a mechanism that already exists in the research platform. Where something is not yet validated, it is labelled.
Research starts with blockchain-native data read directly from chain nodes, not from third-party summaries.
Objects and relationships are reconstructed deterministically, so the same source range produces the same result.
Source identifiers stay attached to analytical results — block height, transaction, and object identity travel with the finding.
Critical findings can be independently re-checked against the source chain rather than accepted on trust.
Research stages may be cryptographically bound through manifest hashes recorded alongside each run.
Operational failures are detected and recovered without silently rewriting evidence; integrity failures stop the pipeline instead.