BR Discovery Engine v0.2: North Star Architecture for Safe Conversational Discovery

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Biocompatible Realism Discovery Engine (BR-DE) v0.2 is an experimental, domain-neutral discovery and falsification framework derived from Biocompatible Realism (BR). It is designed to help transform unconventional, incomplete, speculative, or conventional questions into evidence-classified hypotheses and minimum informative experiments. BR-DE combines viability and constraint analysis, pairwise and higher-order interaction search, feedback and displacement analysis, failure search, an explicit evidence ledger, falsification criteria, and iterative updating. Version 0.2 preserves the frozen 14-stage discovery core developed in v0.1 while adding a North Star and safety architecture for conversational and human-AI use. The framework's North Star is peace: reducing destructive conflict while preserving disagreement, diversity, truth-seeking, and individual agency. Peace is explicitly distinguished from sameness or coerced agreement: unity without uniformity. Version 0.2 adds Life & Agency and Truth gates, anti-harm and anti-coercion safeguards, guidance for handling unusual or speculative claims, a controlled role for humour, and a conversational product architecture intended to make rigorous discovery accessible without requiring users to operate the underlying methodological machinery. BR-DE does not claim to validate Biocompatible Realism, metaphysical or energetic mechanisms, or unconventional claims by compatibility with the framework. It does not claim superiority to existing scientific-discovery methods or AI systems. Such claims require independent comparative validation. The guiding methodological principle is: “Ask the ridiculous question. Demand the boring evidence.” BR-DE is maintained separately from canonical Biocompatible Realism v1.1 (DOI: 10.5281/zenodo.22818247).