Breaking down: "I need to brief a teammate on evidence preserved with a Keryx research receipt. What are the two or three facts I should lead with?"

Identified 3 research target(s) to investigate; these are not established facts

Deep mode: up to 4 paid/cached/public reads plus one bounded gap-expansion pass when needed.

Discovered 21 verified creator source(s) and 4 free public reference(s)

Recalled 60 past runs on this subject — how these sources performed when they were available.

ERC-8004 reputation loaded — composite scores on this subject.

Claim-aware portfolio selected 2/3 positive proposal(s): 2 cached + 0 fresh, predicting 3/3 claim(s) above the evidence floor with $0.000000/$0.015000 fetch USDC reserved.

Free-preview pre-check maps an actionable source to every sub-claim (3/3); paid reading may proceed within the budget.

Agent spend wallet ready: 0x29028Fe1122E17Fe7863A22701e863FE4DaE1aFB (balance sufficient)

Reused cached Keryx Engineering (first-party) — How Keryx pays cited creators (free) — S1

Reused cached Arc Settlement Benchmarks — Measuring x402 settlement latency on Arc (free) — S2

Sub-claim "What evidence is preserved with a Keryx research receipt?": 70% covered by S1 — S1 directly describes the evidence records and ledger: the model proposes evidence records containing a research-question index, source marker, exact quote, and estimated support score; the deterministic evidence ledger checks these against original gathered content; a quotation must occur in the source after whitespace normalization and be 8–240 characters; and the receipt distinguishes settled and pending payment records. This answers the core evidence-preservation question, though it does not enumerate every possible receipt field.

Sub-claim "What are the key facts about how Keryx distributes USDC crea…": 80% covered by S1 — S1 states that citation rewards are separate payments considered after the answer and evidence check; cached content can avoid a new access toll while still qualifying for a citation reward; a paid read does not guarantee a citation reward; the budget is split into access budget and citation pool via a configured citation-pool ratio; the model proposes contribution weights for accepted sources; Keryx resolves weights and allocates the citation pool in integer micro-USDC; only ledger-accepted sources enter allocation; creator payment calls use the source's authoritative payment terms; and unqualified citation pools stay unspent. This directly answers the distribution mechanics, with only minor gaps such as exact formula or timing.

Sub-claim "What are the key facts about how Keryx pays content access t…": 50% covered by S1 — S1 provides partial facts: access tolls are separated from citation rewards, cached content can avoid a new access toll, the research budget includes an access budget, and access tolls already settled remain paid even when no citation qualifies. It does not fully explain the toll-payment procedure, authorization, or settlement mechanics, so this is only a partial answer.

Coverage is adequate for the briefing: S1 directly answers the receipt-evidence and USDC creator-reward distribution questions, and partially answers access-toll payment. The skipped sources concern x402 settlement latency, USDC settlement speed, idempotency, disputes, and unrelated topics; none clearly fills the remaining access-toll procedure gap, so buying more is not justified.

Final check — "What evidence is preserved with a Keryx research receipt?": 90% assessed by S1

Final check — "What are the key facts about how Keryx distributes USDC crea…": 80% assessed by S1

Final check — "What are the key facts about how Keryx pays content access t…": 70% assessed by S1

Final coverage assessment — The supplied passages directly answer what evidence is preserved with a Keryx research receipt and give key facts about citation rewards and access tolls, but the citation-reward and access-toll sub-claims are not the main briefing question and are only partially covered. S1 is first-party engineering documentation and explicitly describes the receipt’s evidence records and ledger validation. S2 is unrelated to Keryx receipts. The assessment does not establish a complete supported answer for every requested part.

Synthesizing a grounded answer from 2 source(s)…

Relevance review returned; only checked excerpts can retain support, and review cannot raise it.

Below support/reward gate — S1 supports claim 1 at 0%: “The receipt distinguishes settled and pending payment records.”

Verified — S1 supports claim 1 at 90%: “It also proposes evidence records containing a research-question index, source marker, exact quote and estimated support score.”

Below support/reward gate — S1 supports claim 1 at 30%: “A proposed quotation must occur in that source after whitespace normalization, and must be between 8 and 240 characters long.”

Verified — S1 supports claim 1 at 40%: “A citation can qualify for a reward only when its source marker appears in the answer, the synthesis declares that marker, and a validated q…”

Verified — S1 supports claim 2 at 60%: “Keryx resolves those weights and allocates the citation pool using integer micro-USDC amounts.”

Verified — S1 supports claim 2 at 70%: “Creator payment calls then use the source's authoritative payment terms; a model-proposed weight does not choose a recipient or change the s…”

Below support/reward gate — S1 supports claim 3 at 30%: “Keryx separates the cost of opening source content from the reward for citing it.”

Verified — S1 supports claim 3 at 50%: “When no citation qualifies, the citation pool stays unspent, while access tolls already settled remain paid.”

Drafted answer citing 1 source(s)

Confidence: Low — the final assessment does not establish a complete supported answer for every requested part.

Keryx Engineering (first-party) contributed 100% - reward $0.015

Settled $0.015 citation reward → 0x6644A7C63C559454e77D5834554DCa3a60fcFDA2 (46aeb37b-3…)

Done. Spent $0.015 across 1 confirmed/simulated payment(s) to creators.