USDC was the bond currency until it wasn't.
The original spec pinned USDC as the bond currency for ReportRegistry. After probing UMA OOv3's Sepolia AddressWhitelist and calling getMinimumBond, we discovered the UMA minimum bond for that USDC token on Sepolia is 400 USDC, completely impractical for a $5-equivalent demo bond. We switched to WETH, which has a minimum bond of 0.002 WETH (a few dollars), and rebuilt the escrow and faucet instructions around it. Mainnet deployment can revisit USDC.

NameStone was replaced by an on-chain contract mid-design.
The original spec used NameStone's off-chain CCIP-Read resolver for all ENS minting; no gas per mint, but a centralized write key held by the project. As the ENS integration deepened, we realised that CCIP-Read off-chain resolvers don't support on-chain writes initiated by smart contracts: ReportRegistry cannot call a NameStone API from inside a Solidity callback. The entire minter service and the off-chain NameStone path were replaced with Lighthouse.sol; an 80-line on-chain contract that calls NameWrapper directly and is the sole write authority for both vessel and verifier subnames. This was a significant mid-flight redesign but resulted in a cleaner and more prize-competitive integration.

spaceTEE doesn't exist yet.
The SpaceComputer architecture documents a spaceTEE inference layer that would let a satellite-hosted TEE predict vessel destinations from imagery. We designed the full integration (plugin methods, inference JSON schema, on-chain attestation) then discovered spaceTEE is on the SpaceComputer roadmap but not yet shipped. Rather than drop the integration, we built the call site honestly: the phare-imager Orbitport plugin exposes the full four-method JSON-RPC interface, but underneath it returns fixture imagery and a rule-based destination prediction that always carries "mocked": true. The KMS signing and on-chain attest() are real; the inference is mock-now-real-later. The plugin is the contribution SpaceComputer can wire to real hardware.