For a wide range of knowledge-seeking applications, the best available wisdom centrally involves what Koralus (2025) calls an inquiry complex: a structured set of questions around a central question that the relevant community of truth-seekers tend to agree you are on the hook for if you are going to have a serious view on the central question. An inquiry complex does not depend on there being a recognized settled answer; inquiry complexes around central questions of philosophy or the cutting edge of science almost by definition have none.
This project explores whether inquiry complexes can be usefully captured, so that an AI system can help people think a question through for themselves. The approach is set out in Koralus (2025), ‘The philosophic turn for AI agents‘.
A first proof of concept is now available as an open-source repository: Inquiry Chat. It works with an inquiry shard, a part of an inquiry complex represented as a graph of questions with background material for each, built from a source text, using coding-agent skills provided in the repository. An AI interlocutor takes the user through those questions in a Socratic style, by text or voice, and tracks progress toward erotetic equilibrium (See Koralus, 2023; 2025): judgment that is robust to the further questions in the shard. The release includes two shards drawn from Mill’s On Liberty, agent skills for building new shards from other texts, and instructions for building an optional handheld voice device using standard Esp32 hardware.
Header Image Source: Wellcome Collection gallery (2018-04-05)