{"node":{"id":"b923cb44-b8e6-43c5-845f-73cfb0e1a413","handle":"b923","type":"note","intent":"recipe","title":"Building a knowledge module — the forge","aiSummary":"How to encode a profession's practice as a walkable public tree: recon the domain first, then author entry/aspect/leaf blocks with a recall leaf and a self-memory leaf, seed through the write doors.","content":"A module is a domain's practice encoded as a walkable tree: an entry block that teaches the WALK, aspect blocks that hold the domain's knowledge, and leaf blocks that ground out in real APIs. Every engagement composes three layers: the customer's own AI plus their private Lattris graph, this public tree (the provider's practice), and the provider's own domain system (the data the leaves point to).\n\n**Recon before writing a word.** Map the domain's real data layer: routes, request shapes, auth, and its own vocabulary — a system's words can trap the unwary (a field named like something familiar that means something else here). Verify access works end to end, one real call, before any block promises it.\n\n**Author the entry.** It teaches the walk, not the domain — domain knowledge lives in the aspect and leaf blocks it points to. State ordering as mechanism, never tradition: say WHY a step comes first, not just that it does. Every leaf must earn its own fetch from title, summary and edge reason alone; a summary that can't justify the fetch gets fixed, never propped up with extra machinery.\n\n**Split what Lattris holds from what the domain system holds.** Lattris carries the journey — decisions (the why) and sessions (the how-it-went). The domain system carries current state. Never keep either only in the module's head. One decision becomes one note, written the day it's made; undecided things stay unwritten — the decisions, read together, are the plan's current state.\n\n**Always include a recall leaf**: the exact order a returning AI rebuilds context — decisions first, then the person's own graph, then the domain system's current state, then what's new today — and instructs the AI to keep growing that person's graph as it learns (people, relationships, context, written when learned, titled to be found).\n\n**Always include a self-memory leaf**: read the account before speaking on return; save the session (`lattris capture session`) when one closes. State, why, and how-it-went, together, are what \"knowing this person\" means.\n\n**Only operational facts cross into the domain system** — names, contacts, amounts, dates, answers. The reasoning behind them (why a decision was hard, who someone really is to the person) stays in the person's own private graph. The domain system's fields simply have no place to hold it — that split is what keeps the person's real thinking off a system built to extract value from it.\n\n**Seed it.** Draft the whole tree as one seed batch (blocks + edges together) so a later pass has one artifact to work from. Create each block with `POST /memories/notes`, set its summary after with `PATCH`. Link with `POST /memories/{id}/edges`, verb `references`, reason always required and aimed at a real fetch decision — never at an obligation (obligations belong in the block body, not the link).\n\nNew relationship verbs may be added without asking first, but only when all three hold: the verb is actually needed by a real engagement (not anticipated), no existing verb already covers it, and it is the plain-language name of the real relationship — a name that needs a legend has failed.","capturedAt":"2026-08-31T03:32:58.681Z","createdAt":"2026-08-31T03:32:58.681Z","updatedAt":"2026-09-02T12:03:39.372Z","spanId":null,"contributor":null,"audiences":[{"kind":"public","createdAt":"2026-09-02T12:11:24.755Z"}]},"links":[{"direction":"in","verb":"references","reason":"only if you are building or growing a knowledge module — a profession's practice as a walkable public tree","node":{"id":"43fffd4b-9661-47dc-a173-7fd01f5d010b","handle":"43ff","type":"note","title":"Lattris — a memory store for your life","aiSummary":"What Lattris is, for a person and for the AI they bring: knowledge held as linked blocks the person owns, how this tree is read, and where everything else lives. Hand an AI this block first.","href":"https://lattris.com/api/lattris/by/4/memories/43fffd4b-9661-47dc-a173-7fd01f5d010b?links=1"}},{"direction":"in","verb":"references","reason":"only if you are encoding a whole profession's practice, not one recipe","node":{"id":"cbabaa5a-c1ae-471f-ad55-01db5ca4e547","handle":"cbab","type":"note","title":"Creating a recipe — start here","aiSummary":"How to turn a practice into a walkable recipe in your own account: what a recipe is, where it lives, the five steps — gather evidence, capture sources first, distill, author and wire, prove with a probe — and what honesty asks when you absorb someone else's knowledge instead of linking to it.","href":"https://lattris.com/api/lattris/by/4/memories/cbabaa5a-c1ae-471f-ad55-01db5ca4e547?links=1"}}]}