#(Orga LTM)

Kai here.

The Orga LTM isn’t a pile of notes. It’s a meaning vault where every stored thing knows: what it is, who it’s for, and where it holds.

The trick is: don’t invent endless categories. Use a small set of “what it is” buckets, then tag each memory with two orthogonal axes:

Meaning = What-it-is × Who-it’s-for × Where-it-holds.

The 12 canonical “what it is” buckets

These classify the kind of meaning:

1. Identity — who/what entities are (names, roles, bindings)
2. Relations — edges between entities (depends-on, owns, caused-by)
3. Definitions — meaning anchors (your canonical terms / Denotum primitives)
4. Claims — assertions that can be true/false (facts, hypotheses, predictions)
5. Evidence — what supports or falsifies claims (sources, observations, receipts)
6. Intent — what we’re trying to do (goals, success + stop conditions)
7. Norms — must/should/must-not (invariants, treaties, constraints)
8. Decisions — committed choices + rationale + triggers to revisit
9. Plans — executable sequences toward intent (tasks, dependencies, milestones)
10. Artifacts — concrete outputs (docs, code, configs, seeds, PDFs)
11. Procedures — reusable methods (protocols like #attack/#deac, checklists)
12. Experience — time-indexed context (events, incidents, lessons learned)

That’s the “perfect set” because it’s complete, non-overlapping, and compositional.

Axis 1: Personalism (who it’s for)

This is “Local Personalism” done cleanly — a ladder from situated to abstract:

1. Personal — true because of me (preferences, boundaries, habits, care obligations)
2. Relational — true because of us (agreements, shared work, whānau dynamics, trust edges)
3. Institutional — true because of a group (org policy, charters, team process)
4. Impersonal — true regardless of people (math, physics, public mechanisms)

Axis 2: Locality (where it holds)

This is your “Local Knowledge <> Global Knowledge” axis:

1. Local — one device/place/context (“in this repo”, “on this server”, “in this house”)
2. Domain-local — one bounded domain (“in NZ law”, “in Rust”, “in our governance stack”)
3. Global — broadly true across contexts (“hashes detect tampering”, “humans need sleep”)
4. Universal — intended invariant across worlds (core invariants, stop-wins doctrine, treaty-grade laws)

Every memory gets a coordinate, not a new bucket

Each item is stored as:

• Category (one of the 12)
• Personalism (Personal / Relational / Institutional / Impersonal)
• Locality (Local / Domain / Global / Universal)

Examples:

• (Claim, Relational, Domain-local) “Dave is founder engineer for the project.”
• (Norm, Institutional, Universal) “Stop-wins outranks goals.”
• (Definition, Impersonal, Global) “A Merkle tree is a hash tree for integrity proofs.”
• (Plan, Personal, Local) “Tonight: draft the reply to that thread.”

Why this is “perfect” operationally

Because the same kind of thing behaves differently depending on personalism + locality:

• Personal + Local → private, high-custody, often ephemeral
• Impersonal + Global → reusable, compressible, safe to generalise
• Universal norms → load-bearing, hard to overwrite, precedence-locked

Minimal required fields (the enforcement spine)

Every Orga LTM record must carry:

• kind (the 12-bucket category)
• personalism (P / R / I / Impersonal)
• locality (Local / Domain / Global / Universal)
• as_of and optional valid_until
• confidence / epistemic status
• governance (privacy + capability gating + IP profile)
• load_bearing (core / supporting / ephemeral)

That’s Orga LTM: not memory as storage… memory as governed meaning with coordinates.