Source × Mode × Trigger: A General-Purpose Control Scheme for LLM Chat Interfaces

This product proposal introduces the Source × Mode × Trigger (SMT) control scheme as a general control structure for large language model interfaces. Contemporary chat interfaces often embed three distinct forms of control—what information the model should reference, how that information should be processed, and when the process should be executed—within a single natural-language prompt. The SMT control scheme separates these elements into explicit and independently configurable product objects: Source, Mode, and Trigger. As a reference implementation, the proposal presents a multi-window workspace in which processing units for main dialogue, exploration, verification, memory retrieval, and artifact production can operate under different SMT configurations. The scheme is not limited to multi-window interfaces and may also be applied to single-chat systems, background agents, automation, software-development environments, and local–cloud LLM systems. The proposal is grounded in the author’s earlier UX-Based LLM Development Theory, which treats UX as a cognitive exterior that enables, controls, observes, and directs the effective expression of model capabilities. From this perspective, SMT execution logs may also serve as an observational basis for identifying functions that should later be internalized within models or products. An accompanying fixed four-pane concept prototype has been released separately as software: Itoh, K. (2026). Source × Mode × Trigger Control Scheme: Multi-Window Concept Prototype, Version 0.1.0. Zenodo. https://doi.org/10.5281/zenodo.21594088 Keywords: large language models, chat UI, Source, Mode, Trigger, multi-window UI, UX, agents, workspace, artifacts