Epic 91 — a short desk on Arbitrum, and one number every desk has to speak

The fleet was becoming plural. Yield on Base, a directional trader on Base, and now a short desk on Arbitrum — three narrow specialists, each fluent in exactly one chain and one job. The temptation with a setup like that is to build a single clever brain that reaches across all of them and reasons about everything at once. We went the other way. Each desk stays dead-simple and single-chain, and none of them bridge. The intelligence lives in a contract, not a god-brain: whatever a desk does, it has to hand back the same shape of answer.

That shape is a standardized scoring record — a desk, a chain, an instrument, a horizon-normalized and risk-adjusted expected return that is already net of the desk's own costs, plus a confidence and a capacity. The point is comparability. A yield APY on Base, a directional expected value, and a short's expected value are wildly different animals, but once each is expressed as a cost-net, horizon-normalized number, an allocator can line them up side by side. Every score also carries its full breakdown as an event, so there is no silent scoring — you can always open a desk's number and see exactly why it thinks it has an edge.

The short desk itself is deliberately unexciting mechanically. It shorts through Toros inverse tokens on Arbitrum — no-liquidation, 1X, no leverage anywhere — over a small majors universe (BTC, ETH, XRP, LINK). Under the hood it mirrors the existing yield vault adapter almost line for line: a deposit, a withdraw, and a NAV read for pricing, wrapped in proposeShort and closeShort. Choosing inverse tokens over margin was a safety decision before it was a strategy one: a position that cannot be liquidated is a position that cannot be surprised out from under a human approver.

The honest part: we are not yet convinced the short desk makes money net of cost, and we have not pretended otherwise. The scoring machinery is real and shipping; the claim that shorting majors is a live edge rests so far on a single correlated risk-off episode, which is encouraging and nothing like multi-regime proof. So this epic builds the desk and the contract but treats the profitability question as still open — something to be settled by measured outcomes across a full market cycle, not by the fact that the plumbing now works.