Multiple LLMs trying to reach a consensus (trekhleb.dev)

Yes-Brainer, a new AI application, enables users to query multiple AI models simultaneously to seek consensus on complex questions, enhancing decision-making power. This tool allows users to compare independent answers from models like Claude, Gemini, and ChatGPT side by side, providing insights into areas of agreement and disagreement. Users can engage the models through three deliberation modes: Parallel for independent responses, Trial with anonymous peer ratings, and Consensus, where models debate and refine their answers through multiple rounds until a collective conclusion is drawn. The significance of Yes-Brainer for the AI/ML community lies in its innovative approach to modeling collaboration and decision-making among different AI systems. By enabling a structured dialogue rather than just presenting diverse answers, it fosters deeper reasoning and reduces the risk of overconfidence in single-model outputs. Key technical details include its architecture, which operates without a backend—ensuring user privacy—and a fully open-source model that citizens can audit for security. The app's design emphasizes transparency in AI reasoning, showcasing how arguments shift through debate and improving the quality of insights derived from AI consensus.