MANTHAN: IDEAS WITH A VISIBLE HISTORY Work architected at Fifth9 Idea-lineage research — This describes an exploration and knowledge-lineage system, not an authority to execute the ideas it produces. Manthan separates generating contributions, relating them, preserving their origins and deciding which deserve further attention. Its lineage record is useful even when the quality of an idea remains unsettled. Selecting consistently useful ideas is a separate question from producing many of them. Can we preserve how an idea develops, including the disagreements that improve it? Preserve how ideas emerge, challenge one another and develop, so later work can use more than a final summary. An idea can change through disagreement, combination and refinement. Recording those relationships makes it possible to inspect how a later proposal depends on earlier assumptions, instead of retaining only the final fluent summary. BUILDING BLOCKS Exploration brief Give a discussion a question and boundaries. Receives: A topic, objective and constraints. Produces: A shared question for contributors. Different perspectives Generate and challenge possible approaches. Receives: The brief and earlier contributions. Produces: Ideas, critiques and refinements. Idea lineage Record how contributions relate to one another. Receives: Origins, contradictions, refinements and combinations. Produces: A connected history of the discussion. Inspection and selection Help people inspect the material and decide what deserves attention. Receives: Contributions, relationships and evaluation criteria. Produces: Candidate ideas and reasons for further investigation. Knowledge handoff Make selected material available with its context. Receives: An idea and its retained lineage. Produces: An inspectable contribution for another person or system. WALKTHROUGH 1. Ask a bounded question Choose the decision or exploration the ideas should help with. The brief gives later reviewers a way to distinguish an interesting tangent from a contribution that addresses the actual question. 2. Generate perspectives Invite different approaches to the question while keeping their assumptions visible. A collection of varied wording is not necessarily a collection of genuinely different ideas. 3. Challenge and develop Record how a contribution derives from, refines, contradicts or combines earlier thinking. These relationships make changes inspectable without assuming that the latest version is automatically the best one. 4. Examine what is useful Inspect the contributions rather than equating volume with value. Assess options in relation to the brief and the evidence still needed. A useful output may be a sharper question or a small experiment, rather than a winning proposal declared by the system. 5. Carry the reasoning forward Hand selected ideas to the next stage with their origins, unresolved questions and assumptions. Keeping the reasoning available helps others challenge it without reconstructing the entire discussion. ILLUSTRATIVE EXAMPLE — Choose a useful first experiment A team is considering several approaches to a new service. Different contributors propose options and challenge their assumptions. The team can trace why an option changed and what remains unresolved when deciding on a first experiment. If a promising option depends on an unsupported assumption, the next step can be an experiment rather than an automatic selection. The uncertainty belongs in the idea record. An option to investigate with its origins and uncertainties available. An illustrative use of idea lineage; it does not assert that automated selection already outperforms human review or a simple baseline. Why these responsibilities are separate The record is useful precisely because it can preserve unresolved disagreement. A selection process may choose an experiment to try, but the existence of a lineage or ranking does not prove an idea is true, practical or worth deploying. SCOPE Multi-persona idea generation, critique and a record of contribution relationships. A viewer and export-oriented work for inspecting and carrying ideas forward. REMAINING QUESTIONS Which ideas improve through challenge? How should selection be compared with a simple baseline? What does a later collaborator need to use an idea responsibly? BOUNDARIES AND EVALUATION Useful thinking: More generated content does not imply more useful thinking. Evaluation: Compare the usefulness of the resulting options and questions with a simpler exploration process. Lineage versus truth: A recorded lineage preserves origins, not the truth of every claim. Evaluation: Trace an important claim back to its contributions and examine their assumptions. Permission to act: Ideas and rankings do not themselves authorise deployment or other external action. Evaluation: Keep the selection of a next experiment separate from permission to carry out consequential action. The aim is thinking that others can question, develop and carry forward without losing its origins. AI Initiative — public architecture guide.