Five stages, one continuous flow.
From raw evidence to published score: each stage appends to the next. Nothing is overwritten. Every stage below carries live counts from the database.
Networks
7
networks live
Every covered network is read continuously — supply, transfers, holder distribution, and on-chain state.
Last activity
—
Collection
7
evidence streams
Raw onchain evidence is collected across every supported network and written to append-only tables. Nothing is discarded.
Last activity
2026-07-07 12:29 UTC
Raw layer
110K
raw observations
Append-only record of exactly what came off the chain — contract reads, transfer events, pool state, oracle values. Never altered.
Last activity
2026-07-07 12:29 UTC
Normalized
—
normalized rows · v1.0
Methodology-driven derivations from the raw layer: annualized yield, liquidity ratios, concentration percentages, capital flow direction.
Last activity
—
Scored & published
54
instruments scored
1
index published
Trust ratings, index levels, and anomaly flags — tagged with the methodology version that produced them. What gets cited, compared, and tracked over time.
Last activity
2026-06-15 09:44 UTC
Append-only architecture
Nothing is altered. Everything is appended. When the methodology changes, the version increments, the change is documented, and historical scores retain their original version tag. A score of 78 under v1.0 and 78 under v2.0 are comparable only within their versions — and that distinction is always visible.
When was each stream last observed?
Each cell shows the age of the most recent observation for that chain × evidence stream pair. Accent green = fresh (under a few hours). Near-white = over a week. Hover for the exact timestamp.
Finding
Freshest: Ethereum / holder snapshots (8h ago). Most recent gap: Avalanche / attestations (16h ago).
Fresh = accent green · >36h = mid · >7d = near-white · Empty = no data
The last ten raw observations.
Assembling on load — a cross-section of the pipeline's most recent ingestion across chains and evidence streams.
| Chain | Stream | Reference | Block time |
|---|
Every published number traces back to source transactions — the chain of evidence is the citation. The raw layer is the append-only foundation; every score, index level, and anomaly flag is computable from it.
30-day activity grid.
Intensity encodes the volume of data the pipeline produced that day, not just whether it ran. Red cells mean every run failed.
Rows · last 24h
3K
data points captured
Rows · 30d
94K
total written
Uptime · 30d
87.1%
successful runs
Cron runs · 30d
480
scheduled executions
Live activity
LiveStreaming across pipeline
No recent activity
How the pipeline earns the score.
Append-only raw layer
Append-only record of exactly what came off the chain — contract reads, transfer events, pool state, oracle values. Never altered.
Every published number traces back to this layer.
Methodology versioning
Every score is tagged with the methodology version that produced it. A score of 78 under v1.0 and 78 under v2.0 are comparable only within their versions.
Change the methodology, recompute this layer. Raw stays intact.
Onchain provenance
Source-kind language on every record: on-chain canonical state, issuer primary publication, independent registry. No fabrications.
The provenance standard governs every persisted fact.
Cleanest available data
We apply the same verification discipline to the pipeline inputs as to the scores themselves. The raw layer is the cleanest source of data you can find — because we have also applied the trace to it.
Contract reads, holder snapshots, published rates, and issuer signals.