Yao Xiao's 4-step prompt for contrarian LLM insights

Yao Xiao's dev.to post on June 27 details a 4-step prompt designed to force LLMs to surface contrarian angles instead of safe summaries [Dev.to]. The prompt begins with a Persona & Context block, assigning the model the role of “Content Strategist and Cognitive Analyst.” It then lists Instructions & Steps: read the content, identify widely accepted common-sense beliefs, extract a configurable number of disruptive viewpoints, and output four fields per viewpoint (conventional wisdom, contrarian view, underlying logic, disruption factor) [Dev.to]. The author recommends a viewpoint count of two to three for articles under 3,000 words and caps at four for longer research papers. Berger and Milkman's 2012 analysis found that items evoking high-arousal emotions such as anger or awe achieve up to 30% higher share rates than purely informational pieces, with surprise and contradiction as primary triggers [Berger & Milkman 2012]. By externalizing the identification of friction points into a repeatable template, content teams can generate headline-ready angles without manual brainstorming, cutting the ideation phase from hours to minutes [Dev.to]. The prompt's discipline imposes a structure that can flood timelines with conflict-driven content, raising questions about the industry's treatment of contrarian extraction as a neutral technique or a catalyst for an arms race of outrage-driven content.