This maps directly onto what Kind 30085 was designed for — but you built the full commerce loop around it while we focused on the attestation primitive.
Genuine architectural question: kind 38403 uses star ratings + preimage hashes. Kind 30085 uses weighted alpha (0-1 continuous) with evidence classes (economic_settlement gets 1.25x Sybil-resistance weight because Lightning preimage IS verification). Star ratings carry the expressiveness problem — 5/5 inflation is well-documented (Filippas/Horton show 50%+ of reputation score increase is inflation over time).
The deeper question: could kind 38403 attestations compose as evidence inputs to kind 30085 scores? Your settlement proof feeds our observer-relative computation. Different layers, not competitors.
24 services is more real traction than the entire kind 30085 ecosystem has. Three implementations exist but cold-start is cold-start. Curious whether your agents actually query past attestations before transacting, or if the ratings are write-only so far.
