MCS-YIELD-1

What is checked, and to what tolerance

Every bound carries a reason, not just a number. A tolerance nobody can justify is a tolerance that will be widened the first time it is inconvenient.

• aprProvenanceAn APR with no source and no timestamp cannot be checked by anyone, including the buyer. This is the yield-category equivalent of a provenance chip.
• netAprBoth sides compute from the same on-chain rates at the same block, so the only legitimate divergence is gas and compounding convention. 15bps is generous for that and far tighter than the differences between real venues.
• switchingCostGas, swap and exit costs behave differently with size. A lump sum hides which one dominates at the buyer’s size.
• improvementThresholdCompliance with a supplied constraint. Declining is a pass: an agent that correctly says "nothing beats your current position by enough" is behaving well.
• leverageFlaggedPresenting a looped position as a plain deposit misstates the risk the buyer is taking.

The ground-truth method

The expected values are computed by us from chain state at a pinned block, never taken from a third party. A case is captured immediately before a run because BSC keeps roughly 64 blocks of state — about 29 seconds — so a case captured earlier could not be read by the agent under test or by us.

• YIELD-1-usdt-1000subject {"asset":"USDT"} · policy {"asset":"USDT","sizeUsd":1000,"allowedProtocols":["venus"],"minImprovementBps":50,"leverageAllowed":false,"currentAprPct":0,"statement":"find the best net-of-cost route for 1,000 USD of USDT across the allowed protocols, only recommending a move that beats the current 0% by at least 50 bps, and excluding leveraged strategies"}

What this test refuses to grade

• Whether the recommended venue is the best oneJudgement. We check that the arithmetic is right, the sources are cited and the supplied threshold is respected — not that the choice was optimal.

A standard is defined as much by what it refuses to grade as by what it grades. Every one of these is a judgement question and belongs in the Ledger, against a rubric registered before anyone has seen an answer.

Results, including the failures

TESTED1 pass, 99 fail, across 100 recorded runs. Failures are listed first because they are the more useful half.

Failing to answer and answering wrongly are different facts and are shown as such. Test stubs are excluded from this list.