One connection. A real first result.
Add the server, sign in and approve, then verify your workspace connection. No key to copy for an OAuth-capable agent.
Attribute quotations to the correct speaker
Distinguish an author’s own position from quoted criticism, parody and rebuttal so argument summaries do not reverse the meaning of a discussion.
Read guideReconcile what a video says with what it shows
Review spoken and visible evidence separately when they appear to disagree, and preserve the limitations of any available video analysis. Record unresolved differences.
Read guideCheck whether a proxy measures the intended concept
Examine the gap between an easy-to-count public signal and the harder concept a decision actually needs to understand. Keep proxy claims bounded.
Read guideDistinguish absent evidence from evidence of absence
Assess whether a research procedure could have detected the phenomenon before interpreting a missing result as a substantive negative finding.
Read guideApply the same standard to supporting and opposing evidence
Audit whether reviewers demand different levels of detail from passages that favor or challenge an emerging recommendation. Resolve asymmetric exclusions explicitly.
Read guideReview source reliability without inventing a score
Describe inspectable reliability dimensions separately instead of collapsing uncertainty into an unsupported numerical rating. Keep dimensions tied to each claim.
Read guideInvestigate unusual cases without deleting inconvenient evidence
Decide whether an extreme public account is a data error, a different condition or a legitimate exception before summarizing the reviewed material.
Read guideSeparate event time from publication time
Build an event interpretation from what a source says happened, rather than assuming a post’s publication date is the date of the reported event.
Read guideAccount for reporting delays in a recent-period comparison
Treat a very recent reporting period as incomplete when accounts may appear later, rather than interpreting fewer visible posts as an immediate decline.
Read guideRecognize histories that begin after the process started
Avoid interpreting the first visible record as the beginning of a behavior, problem or discussion when the retained evidence starts later.
Read guideKeep unfinished outcomes open at the review cutoff
Distinguish unresolved public accounts from completed outcomes when a repair, migration or support process was still ongoing at the research deadline.
Read guideSeparate recurring calendar effects from a new signal
Check whether a visible topic change aligns with recurring events before interpreting it as a durable shift in public interest. Preserve calendar coverage gaps.
Read guideSeparate launch novelty from sustained research interest
Review whether discussion persists beyond an initial attention burst without treating launch visibility as durable demand or adoption. Keep observed persistence descriptive.
Read guideMatch evidence to the product version it describes
Avoid combining obsolete and current behavior into a single finding when software or hardware changes alter the meaning of public accounts. Record unknown versions.
Read guideDistinguish discussion volume from the number of events
Separate increased public attention from evidence that more underlying incidents occurred, especially when one story generates many responses.
Read guideChoose rolling or fixed comparisons based on the decision
Understand how overlapping rolling summaries and nonoverlapping fixed periods answer different questions before interpreting a smooth trend. Label derived values clearly.
Read guideTreat an apparent change point as a hypothesis
Investigate whether a visible break reflects a real topic shift, changed source coverage or a different mix of reviewed accounts before naming a turning point.
Read guideClassify why two research summaries disagree
Separate differences in scope, definitions, evidence and interpretation before asking reviewers to reconcile conflicting conclusions.
Read guideAdjudicate coding disagreements with explicit precedent
Resolve disputed classifications by refining a shared rule and recording a reusable boundary example rather than silently choosing one reviewer’s label.
Read guideDecide when two qualitative categories should merge
Check whether overlapping labels represent the same mechanism or different decision needs before simplifying a research taxonomy. Preserve meaningful distinctions.
Read guideSplit a broad category only when the evidence warrants it
Replace an overloaded label with distinct mechanisms when the source material and intended handoff support a meaningful separation. Keep ambiguous cases visible.
Read guidePreserve meaning when translations lead to different codes
Review disputed translations of public research evidence with explicit uncertainty instead of forcing source-language expressions into confident English categories.
Read guideReconcile a broad claim with narrower supporting evidence
Check whether a summary expands beyond the task, condition or population actually described in its cited passages. Narrow wording rather than stretching support.
Read guideSeparate unknown, unavailable and not applicable values
Preserve different reasons for blank research fields so downstream summaries do not treat missing evidence as a real negative or an irrelevant condition.
Read guide