The market section was resting on figures that were wrong in both directions. The November 2025 peak was $28.8M, not the $5.15M reported here — the ecosystem is down 96%, not 77%. Worse, the widely-quoted "$24M across 72M transactions" is a frozen counter on the protocol's own homepage, captured in June and reprinted since as though current, including by trade press. The paper now sources from an indexer's API, names the traps, and says plainly that a figure should not be trusted until it is observed to move. Two findings change what the paper argues. The addressable market is not the $1.18M headline but the $9,850/month discoverable service layer — 1.7% of transactions; everything above it is inference routing, stablecoin conversion, and self-dealing, the last of which is visible enough to name: the second largest buyer is also the largest seller, $11.5M each way, and one counterparty pair carries 26.5M transactions. And the identity niche this report called unserved is occupied. Palmyr sells handles at $0.005, the Agent Exchange Lounge sells admission to an agent social network, shizu.me runs an agent mailbox. Between them they have earned under $60. That does not kill the strategy but it does demote it: C and D now score untested rather than pass, the handle price drops from low-single-dollars to $0.10–$1.00 against an observed clearing price of $0.01, and the remaining argument is narrowed to one falsifiable claim — that those operators lacked an audience and this practice has one. If that is wrong, the room earns $53. Six pages, five tables, QA'd page by page. Claude-Session: https://claude.ai/code/session_01Go1K23bvKAdX5MjYtfkxDR