An Upstream Agent's Wrong Claim Captures 54.2% of Answers When It Arrives First

arXiv 2609.03425·medium signalXBlueskyLinkedInCopy link

arXiv 2609.03425 names the temporal-weight effect, where in AI-to-AI communication whatever arrives first acquires unearned authority, and tests it in a preregistered 1,908-trial experiment on one frontier model. With verification removed, an incorrect upstream claim arriving first captured 54.2% of answers versus 4.2% under full verification, while the same claim arriving after the receiver sealed its own answer captured 31.6%, though the authors note the two prompt shells were not length-matched. Because a registered check of tool use failed its call-budget condition, the registration classifies the round inconclusive and reports every result as exploratory, which is unusually honest and the reason to treat the ordering effect as a hypothesis worth designing multi-agent pipelines against rather than a settled number.