A practical guide to deterministic evidence briefs in TEMIRIN using configured records, explicit market identity, current controls, and human review.
July 5, 2026 · 3 min read
An evidence brief is useful when it narrows a review without hiding the underlying record. Begin with the exact Polymarket question, the outcome being considered, the close time, and the venue's stated resolution source. Then state the research question in plain language: which new fact was observed, which part of the contract could it affect, and what remains uncertain? This framing prevents a concise brief from drifting into a generic news summary that has little connection to settlement.
TEMIRIN briefs organize stored source fields deterministically. They can present source tier, reliability, relevance, direction, freshness, canonical provenance, and tagged market context. Those fields are not generated predictions. The reviewer should be able to move from every summary line to the saved source record and its timestamps. If the evidence lacks an explicit supported market identifier, keep it as research material instead of forcing it into a watchlist opportunity.
For an RSS or HTTPS web item, retain the publisher, canonical URL, supplied publication time when present, observation time, stored excerpt, and ingestion state. Canonicalization and deduplication can suppress repeated copies while preserving enough information to explain what was accepted. Provider-configured Telegram and X readers retain the references available through their providers. A brief should never erase these differences by presenting every channel as if it offered the same metadata.
Freshness needs two clocks. Publication time describes when the source says the item appeared; observation time describes when the workspace received it. A corrected page may keep its original publication time while its content changes later. A provider failure may make a seemingly recent item the latest available but not the latest published. Showing both clocks helps the reviewer distinguish evidence age from collection health and document any limitation before recording a conclusion.
Reliability can describe a configured source, relevance can describe its connection to the tracked topic, direction can capture the tagged interpretation, and freshness can describe timing. Review each input before relying on the combined confidence context. A high value cannot repair a wrong market ID, ambiguous resolution rule, stale public price, or weak source. A low value may simply signal incomplete metadata that requires manual inspection rather than a negative view of the market outcome.
Compare the brief with the original document and with any clearly independent primary evidence. Do not count syndication, screenshots, or copied posts as separate support. Record disagreement as a reviewer note with citations instead of inventing a hidden classification. The brief should make the path of reasoning inspectable: what the system stored, what deterministic fields were shown, what the reviewer verified, and which uncertainty remained at decision time.
Once evidence is explicitly linked to a watchlist market, compare the stored trigger price with the latest available public price and quote timestamp. Bid, ask, spread, and visible depth can describe the current book view, but they are observations rather than fill promises. Save the exact time used in the brief and reopen the venue question before making a decision. A material price move may change the review even when the source itself has not changed.
Record approve or reject inside the authenticated workspace, with the actor, time, and note kept distinct from alert delivery. If the user later enters a paper order, the simulated request can be checked against the per-order limit, rolling twenty-four-hour paper total, per-market exposure, and portfolio concentration. It remains separate from the evidence brief and from every real transaction. This separation lets later journal review compare information, decision, and paper outcome without rewriting what was known.
A brief can present stored source tier, reliability, relevance, direction, freshness, tags, canonical reference, observation time, and explicit market identity while keeping the underlying evidence available for human inspection.
Each timestamp describes a different event. The publisher supplies publication time when available, TEMIRIN records collection time, and the workspace records later human actions, making the evidence sequence easier to audit.
It is a separate user-created simulation linked through known record identifiers. Its direction, notional, limit price, and server-check results remain distinct from source quality, market price observations, and the reviewer’s note.
