A practical guide to deterministic source-reliability context in TEMIRIN using configured records, explicit market identity, current controls, and human review.
August 3, 2026 · 4 min read
A source-reliability value should summarize documented provenance and collection behavior, not decide whether a Polymarket outcome will occur. Define the rubric in plain language before assigning values. Useful criteria include canonical authority, identifiable ownership, correction practice, stable access, observation freshness, and whether the source publishes the record named by a market’s resolution rules. Preserve an unknown state when history is too thin for a supported assessment.
Give each criterion an evidence requirement. “Official” should point to an owned domain or authenticated provider record; “fresh” should reference an observation time and expected update pattern; “corrected” should retain the earlier and later versions. A reviewer should be able to reconstruct the value without relying on reputation alone. Document the rubric version so later changes do not silently alter what an old score meant.
Decide how the rubric handles a source that is authoritative for one record type but weak for another. A regulator may be primary for a filing and irrelevant for a network metric; an exchange may be authoritative for its own listing status but not for a protocol decision. Scope the assessment to the configured source purpose instead of assigning one universal reputation value.
Review configured RSS, web, Telegram, or X sources through their own records: canonical reference, successful and failed observations, duplicate behavior, supplied publication time, and documented corrections. A source that is authoritative but temporarily unavailable has a different problem from a reachable page with unclear ownership. Keep collection health and provenance visible as separate inputs so support can choose the right response.
Repeated articles do not become independent reliable sources when they trace back to one announcement. Preserve each collected path for audit, identify the original authority, and deduplicate the evidence relationship. Outbound email, Telegram, Slack, or webhook delivery history is also separate: a failed notification says something about transport, not about the reliability of the underlying source record.
Reliability does not create a relationship to a market. A source item appears beside a watchlist-derived review record only when the supported explicit market identifier is present and the reviewer has checked the contract. An authoritative publisher can still be irrelevant to a particular threshold, date window, or outcome. Store the relevance note and market identity independently from the source’s provenance value.
Trigger and current Polymarket prices can show how the market changed around evidence, but movement should not raise or lower reliability. The configured public-wallet snapshot and policy state likewise belong to risk review, not publisher assessment. Keeping price, portfolio, policy, decision, and source dimensions separate prevents a favorable paper mark or preferred thesis from changing how the same source is described.
Sample sources across high, low, and unknown values and rebuild each assessment from stored facts. Include a canonical primary source, a syndicated summary, a corrected record, a prolonged collection failure, and a new source with little history. Review whether two people using the written rubric reach a comparable result and record why they disagree. The exercise tests clarity rather than predicting market direction.
When criteria change, publish a new rubric version and apply it prospectively. Preserve earlier values with their original definition so historical reports do not compare rewritten assessments with current observations. Review decisions and paper records may link back to the source value available then, but their per-order, rolling twenty-four-hour, per-market exposure, and portfolio-concentration results remain separate simulation evidence.
Schedule reassessment from observed source behavior rather than from a favorable or unfavorable market outcome. A correction, ownership change, repeated access failure, or altered publication process can justify a new review. Record the specific event and reviewer so another teammate can understand why the value changed without reading the later market result into the source history.
It should use documented provenance and operations criteria such as canonical authority, ownership, correction behavior, freshness, access health, and an explicit unknown state.
Not when they trace back to the same origin. Preserve the collected paths, identify the original authority, and avoid treating syndication volume as separate confirmation.
No. Evaluate the source against the declared provenance rubric. Market-price observations, portfolio context, decisions, and paper records remain separate review dimensions.
