Knowledge is a graph

Cooked up the night before with Julia and formally dated on camera, “quería marcar el inicio de esta nueva idea, hoy, 16 de agosto” ▸ 26:56, the next project reimagines the chat interface for learning. Instead of a scrolling thread, each conversation is a canvas; instead of one long answer, a graph.

The founding example is Marvel lore ▸ 22:20: ask to learn it and the app sketches three nodes, villains, heroes, antiheroes, each with a brief description. Ask a node about Thor and the graph grows outward from the essential toward the specific ▸ 23:00. Julia’s mechanic makes exploration frictionless: any key term in an answer, Asgard, is clickable, and the click spawns Asgard’s own bubble ▸ 25:45. Two more features complete the spec: an “I didn’t understand” button that decomposes a concept into the sub-terms needed to grasp it ▸ 23:48, prerequisites made navigable, and, riding Gemini’s giant context window, the option to drop in an entire book and have it unfold into an explorable chapter tree ▸ 24:08.

la respuesta deja de ser un texto y se vuelve un territorio →

The idea sits at the confluence of the diary’s obsessions: the eyes-that-focus instinct of the old foveated-vision research (seed 53-1), Pipo’s explorable worlds, Zenota’s node trees, and the midnight Neo4j sketch from seed 66-1, all converging on one claim, the catalog’s aphorism: knowledge is a graph, and learning is walking it.

And it will be built under entry 73-2’s new rules, minimal, no landing page, for the brave, because, as he admits while listing Sanfanson, Zenota, and Divo: “yo al final termino haciendo todo esto para mí” ▸ 26:09. The first brave user is always the same person…