Dataset selection

The May 2026 Amaru Treasury lattice was assembled from the operator's named May batch plus a one-hop transaction-input closure.

Step 1 - start from the 30 named seed transactions

The seed set is the May 2026 operating batch used by the original SPARQL presentation:

source	count
disbursements	3
reorganize transactions	5
swap-order transactions	20
swap cancel	1
scoop dive	1
total seeds	30

Seed txids are included in selections.txt. They are also preserved in the local Amaru Treasury transaction archive under transactions/2026/ in the source operator repository.

Step 2 - fetch closure parents from transaction CBOR

For each seed transaction, the closure walk fetches parent transaction CBOR for every consumed input at depth 1. The reproduce pipe in README.md invokes cq-rdf body --provider blockfrost (or any other supported provider) to do the CBOR pull; the only indexer endpoint used is the transaction-CBOR fetch (no address, UTxO, input, or output convenience APIs).

The reproduced selection list was cross-checked from two sources: Koios tx_utxos for normal consumed-input parents, and the local operator archive's inputs/<parent-txid>.cbor files for reference and collateral parents that are not returned by that Koios endpoint.

Step 3 - write one txid per CBOR

The final lattice contains 101 transaction CBORs:

source	count
seeds	30
normal consumed-input parents	63
reference/collateral parents from local CBOR archive	8
total in lattice	101

The full enumeration is selections.txt, one txid per line. Return to the presentation.