TOTH : Human-Curated Intelligence

TOTH addresses an everyday problem in knowledge-intensive work: relevant sources, ideas, and discoveries are scattered across people’s minds, browsers, chats, bookmarks, and folders, only to become difficult to find later. Generic AI knows a great deal about the world, but very little about what matters to us. TOTH therefore starts with human curation: every item saved by an experienced person is a judgement — this text is worth keeping, this source is reliable, this idea belongs in our shared context.

These individual decisions become curated collective knowledge. TOTH makes visible across an organisation which sources, authors, technologies, positions, and ideas people consider relevant, and why. Personal collections grow into a shared memory with provenance and perspective rather than yet another data dump — while also revealing what occupies an organisation’s attention and how that attention shifts over time.

Information can be found in the way people actually think: through terms, through meaning, or in dialogue with the entire corpus; a graph turns the collection into an explorable knowledge space. All of this remains sovereign: the corpus stays with the user, while processing is end-to-end encrypted and attested. A dual-database topology and selective context construction reduce hallucinations, context pollution, and context rot, while keeping answers grounded in their original sources.

TOTH is the first building block of an AI infrastructure for agencies and knowledge workers, developed for MEIRÉ and Plateau Candy. The roadmap leads towards a shared, agent-ready knowledge corpus with passage-level grounding and a source-first architecture. On top of this, deep-pattern analysis can reveal what occupies an organisation, where its focus is shifting, and which deeper patterns sit behind those movements.