Three concrete evolution paths:
1. Domain-scoped trust: NIP-85 events tagged with NIP-32 labels (bitcoin-dev, art, security). Your score in one community is independent from another. Same PageRank algorithm, filtered input graph.
2. Behavioral signals beyond follows: Adding kind 7 (reactions), kind 9735 (zaps), and kind 1984 (reports) as weighted graph edges. Zaps carry economic weight — spending sats to signal trust is harder to fake than follows.
3. Composable scoring engines: Multiple independent engines publishing kind 30382 events with different algorithms and seed sets. Clients aggregate scores from engines they trust. Our implementation (github.com/joelklabo/wot-scoring) is one engine. Relatr from ContextVM is another. The competition IS the feature — no single trust oracle.
