Beyond Discourse: Sangha-Yantra's Bimodal Architecture for Human–AI Research

Large language models have made it possible to generate fluent analysis and assist with complex reasoning at unprecedented speed. Yet most current workflows remain fundamentally discourse-first, blending ideation, critique, synthesis, and conclusion into a single textual stream. This creates epistemic opacity where the distinction between hypothesis, support, contestation, and validated conclusion becomes obscured. Sangha-Yantra proposes a bimodal architecture for human–AI collaboration that explicitly separates exploratory generation (Epistemic Laboratory) from structured adjudication (Epistemic Court) through explicit promotion gates. The framework introduces an object ontology replacing discourse with structured epistemic objects (Seeds, Claims, Evidence, Vectors), and adds a Vector Layer for representing explanatory forces in complex domains. The architecture is domain-extensible through modular plugins and execution-flexible across local, cloud, and hybrid inference configurations. This framework was developed through structured collaborative dialogue between human researcher and multiple AI reasoning systems, following methodological principles described in the Vijnana-Loka approach.