Show HN: Isnad – A Python framework that uses 1,200-yearold Islamic logic for AI (medium.com)

A new open-source Python framework called ISNAD has been introduced, inspired by a 1,200-year-old system from Islamic hadith scholarship that addresses issues of trust in multi-agent AI systems. The challenge faced by many current AI setups is determining which agent in a chain introduced an error when retrieving and processing information—often leading to a loss of traceable trust. ISNAD aims to tackle this problem by modeling every AI transformation as part of a transmission chain, where the reliability of each agent is documented and influences the overall trust in the final claim. The significance of ISNAD lies in its innovative approach to trust provenance, moving beyond traditional methods by implementing a weakest-link evaluation system, where trust is bounded by the lowest reliability grade of the agents involved. By allowing AI researchers and practitioners to analyze how trust propagates through chains of transformations, ISNAD encourages a paradigm shift in evaluating and improving multi-agent AI systems' reliability. The framework is already operational, featuring a public research paper, extensive test coverage, and initial experimentation, paving the way for future innovations in trustworthy AI and providing a foundation for further validation and integration with tools like LangChain.