Data flow

Where every Atomicals answer comes from. Universe-operated Bitcoin Core produces blocks. Atomicals ElectrumX reads them and applies the protocol rules, which is where protocol interpretation happens. A Universe index scans that provider and publishes an immutable generation as a materialised read model. Products read the index through a public API. Each hop can be behind the one before it, so every answer carries the generation identifier and indexed height it was computed at. No third-party blockchain data provider sits in any production path.

Where every Atomicals answer comes from

Universe-operated Bitcoin Core produces blocks. Atomicals ElectrumX reads them and applies the protocol rules, which is where protocol interpretation happens. A Universe index scans that provider and publishes an immutable generation as a materialised read model. Products read the index through a public API. Each hop can be behind the one before it, so every answer carries the generation identifier and indexed height it was computed at. No third-party blockchain data provider sits in any production path.

The hops

1. Bitcoin Core produces blocks. Universe operates the node.
2. Atomicals ElectrumX reads blocks and applies the protocol rules. This is where protocol interpretation happens.
3. A Universe index scans the provider, builds a materialised read model, and publishes it as an immutable generation.
4. A product reads the index through a public API.

Each hop can be behind the one before it. A consumer must know which position each answer came from.

What can be stale, per hop

Hop	Can be stale because
Bitcoin Core	Normal block propagation
ElectrumX	Still processing a block, or a reorg
Index	Scan in progress, or generation build in progress
Product cache	Cached an answer without its generation

The last row is the one under your control, and the one most often responsible for a wrong screen.

What the ARC-20 side guarantees

The scan is all or nothing. Holder rows are aggregated by address and must sum exactly to circulating supply. A mismatch, a null address, a truncated page sequence, a changing declared total, or a changing chain tip aborts the entire poll. No partial authoritative snapshot is ever committed.

That is why an answer is either complete or absent, rather than plausible and wrong.

What the NFT and Realm side guarantees

NFT, Realm, and Subrealm rows share one provider scan, one generation, one checkpoint, and one active pointer. Publication requires exact final declared source-row coverage and rejects a count shrink or an incomplete page.

A generation that spans more than one provider tip is published explicitly as mixed and stale, without a single indexed height, rather than pretending to be consistent.

What to record with every answer

Field	Why
Generation identifier	So you can detect a rebuild
Indexed height	So you know the chain position
Source revision	So a behavior difference is explainable
Request identifier	So a support question is traceable
Stale and mixed-tip flags	So you know whether to act on it

Storing an Atomicals result without those five makes any later disagreement unexplainable.

Source

Consistency and reorgs and readiness and freshness.