Mohit Yadav Presentation
AI: Summary
The session centred on a guest presentation by Mohit Yadav exploring how meaning arises through dynamic assemblages of known and unknown elements during encounters with the world, arguing that current tools fail to intervene at the moment of interaction and that a new category of “navigational systems” is needed to bring latent personal resources into active contact with present experience. The ensuing discussion wove together knowledge graph visualization, the cognitive role of fuzziness and analogy, the spectrum from precise definition to deliberate conceptual blurring, and the limits of graph-and-node metaphors when modelling genuine intellectual work.
Mohit framed cognition as embodied, distributed, and path-dependent, drawing on process ontology, distributed cognition, and complex systems theory. The progression from orality to literacy to the internet was cast not merely as a change of medium but as a restructuring of cognitive capability itself—each medium reshaping the assemblages through which meaning can emerge. The core proposal was that a distinct category of system is needed: not search, not recommendation, but navigational systems that intervene during the moment of encounter, surfacing latent personal resources (bookmarks, notes, prior readings) so they can participate in the meaning-making process as it unfolds. The theoretical underpinning drew on the linguist Lera Boroditsky’s work showing that language actively structures perception, the art historian B. N. Goswami’s notion of “making the invisible visible,” the vision neurobiologist Margaret Livingstone’s argument that perception is information processing rather than image transmission, and the artist Robert Irwin’s shift from object to context. The presenter argued that current app-centric, siloed digital environments prevent the kind of cross-pollination that happens naturally in cognition, and that spatial, non-linear interfaces—rather than desktop or screen metaphors—are better suited to enabling this unfolded interactivity.
In discussion, Tom Haymes demonstrated early prototypes of a Knowledge Navigator tool that uses a large language model to decompose documents into graphical node-and-connection maps, with filtering and iterative re-generation so that each run yields a different perceptual angle on the same text. The tool was described as a “perception engine” designed to help users think through complex ideas by fusing multiple documents and revealing non-obvious connections—such as parallels between military history and technology adoption.
A key tension emerged around the nature of concepts and the adequacy of node-graph representations. One participant argued that graphs and trees only scratch the surface of genuine intellectual work, which operates across a spectrum from strict one-to-one sign-referent correspondence to the deliberate fuzziness of metaphor, analogy, and poetic language. The distinction between definition (a fixed object) and differentiation (a contextual, layered process) was proposed as fundamental: the same concept—”intellect” versus “cognition,” for instance—carries entirely different theoretical commitments and histories depending on the tradition invoked, and a serious interface must let the user control which distinctions matter and which can be collapsed. This participant reported building software with Claude Code that rewrites text like a diffusion engine, rerouting an entire document along a user-specified map of differentiations rather than definitions—a practical experiment in making conceptual topology computationally tractable.
The session converged on the observation that current AI tools excel at focused research when given clear instructions but remain poor at noticing what is adjacent—the lateral, serendipitous connections that often prove most generative. Broadening working memory through interface design, while leaving the formation of fuzzy conceptual relationships to human cognition, was proposed as the productive division of labour between tool and mind.
Brief Author visionOS Update
Documents can now be shown in sections along with defined concepts:
Origami Text Update
The pitch for Origami Text has been updated.