Bringing material in

Every one of the six tools depends on the same raw ingredient: material that a person actually cared enough about to keep. PDFs, notes, saved pages, transcripts, exported chats with other assistants — the ordinary residue of paying attention to something over time. The knowledge base’s only job is to take that residue in without losing it or garbling it.

“Without losing it” sounds obvious until you have watched a document upload silently truncate a scanned page, or a note lose its formatting and, with it, the distinction between a heading and a stray sentence. The base treats ingestion as a job with a visible outcome: a document either finishes, in a state you can inspect, or it is flagged, in a state you can also inspect. Nothing disappears quietly.

Fig. 1.1 — A document has exactly two honest outcomes: searchable, or visibly flagged. Never a silent third option.

This is a modest ambition, described plainly, and that is deliberate. A knowledge base that merely stores files is not a hard problem. A knowledge base that a person can trust enough to stop keeping a second copy elsewhere is a different and much harder promise — and it is the promise this tool set out to keep before anything more ambitious was built on top of it.