Glossary

Knowledge Base · scientific short definitions · Edition 2026-07-12

Notation: italics = informal gloss · code = common technical spelling.

Offline Wikipedia index

Snapshots live under .gitbook/assets/snapshots/. Each major term below points to a live Wikipedia URL; when a figure is present in a chapter, prefer that chapter’s Offline snapshots section.

---

A

Affordance
Relation between an agent’s action capabilities and features of the environment that makes an action possible (Gibson / Norman).
→ https://en.wikipedia.org/wiki/Affordance

Algorithm
Finite, unambiguous procedure transforming inputs to outputs.
→ https://en.wikipedia.org/wiki/Algorithm

Alignment (ML)
Steering model behaviour toward human-intended goals and constraints.
→ https://en.wikipedia.org/wiki/AI_alignment

Anthropomorphism
Attribution of human traits to non-human entities. Distinct from sociomorphing.
→ https://en.wikipedia.org/wiki/Anthropomorphism

Autonomy (personality)
Orientation toward independence and self-definition; contrast sociotropy.
→ related: https://en.wikipedia.org/wiki/Sociotropy

---

B–C

Base direction
The initial text direction of a string, ltr or rtl, stored as a third part of an RDF 1.2 literal beside text and language: "الأم تحكي قصة"@ar--rtl; JSON-LD 1.1 writes "@direction": "rtl". RDF 1.1 has no slot for it.
→ IEML × RDF × Situation Monitor

Bi-temporal edge
A fact edge with two time axes: valid time (valid_at, invalid_at: when the fact held in the world) and transaction time (created_at, expired_at: when the system believed it). A contradicting fact closes the old edge instead of deleting it (Graphiti).
→ Agent memory and world-model repos

Blast radius
Growth a plan reports before apply. In the silent-policy transcript the triple count stays 0 while the rule table adds new consequences. On this desk the count that gates a write is article ids and org file ids.
→ Concepts governance

Blockchain
Replicated append-only ledger maintained by a consensus protocol.
→ https://en.wikipedia.org/wiki/Blockchain

Cache
Fast intermediate store that exploits locality of reference.
→ https://en.wikipedia.org/wiki/Cache_(computing)

Certificate (plan)
Record that names a plan, the rule-table hash that ran, and the closure diff that check computed. A changelog or a cache write of a bare edge list is a different object.
→ Concepts governance

Category theory
Abstract study of objects and structure-preserving maps (morphisms); functors map between categories.
→ https://en.wikipedia.org/wiki/Category_theory

Channel capacity
Supremum of rates for reliable communication over a noisy channel (Shannon).
→ https://en.wikipedia.org/wiki/Channel_capacity

CIA triad
Confidentiality, Integrity, Availability — classical security goals.
→ https://en.wikipedia.org/wiki/Information_security

Cipher
Algorithm for encryption/decryption under a key.
→ https://en.wikipedia.org/wiki/Cipher

Collective intelligence
Intelligence appearing at group / network scale beyond individual members.
→ https://en.wikipedia.org/wiki/Collective_intelligence

Composition (arts)
Arrangement of formal elements guiding attention and meaning.
→ https://en.wikipedia.org/wiki/Composition_(visual_arts)

Computer-use agent
An agent that observes a graphical interface (accessibility tree, DOM, screenshot) and acts on it with clicks and keystrokes, usually inside a VM or container sandbox (Cua).
→ Agent memory and world-model repos

Conservativity
A change alters no consequence, under a named Horn rule table, over the names the store already uses. As a hard gate, a standing query that would gain ids blocks the insert.
→ Concepts governance

Consensus (distributed)
Agreement on a value or log among processes despite faults.
→ https://en.wikipedia.org/wiki/Consensus_(computer_science)

Cryptography
Study of secure communication and computation under adversaries.
→ https://en.wikipedia.org/wiki/Cryptography

---

D–F

Diagnosis (medicine)
Assignment of a disease label given signs, symptoms, and tests.
→ https://en.wikipedia.org/wiki/Medical_diagnosis

Entity resolution
Deciding whether two mentions name the same entity, then merging them under one id. Graphiti runs exact name, then entropy-gated MinHash/LSH, then an LLM judge; screenpipe resolves only on normalized identifiers and returns unique, ambiguous, or not found.
→ Agent memory and world-model repos

Entropy (information)
$H(X) = -\sum_x p(x)\log p(x)$ — average uncertainty of a discrete source.
→ https://en.wikipedia.org/wiki/Entropy_(information_theory)

Episode (agent memory)
One unit of raw input (a message, a JSON record, a text, or a fact triple) with a reference time. Every derived node and edge keeps the ids of the episodes it came from, so an episode is the provenance record (Graphiti).
→ Agent memory and world-model repos

Epistemology
Theory of knowledge: justification, belief, and truth.
→ https://en.wikipedia.org/wiki/Epistemology

Explicit knowledge
Codified, transmissible knowledge (documents, code, schemas).
→ https://en.wikipedia.org/wiki/Explicit_knowledge

Field (algebra)
Set with $+$ and $\cdot$ forming a commutative division ring (e.g. $\mathbb{R}$).
→ https://en.wikipedia.org/wiki/Field_(mathematics)

Functor
Map between categories preserving composition and identities.
→ https://en.wikipedia.org/wiki/Functor

---

G–I

Generalization (ML)
Performance of a learned function on unseen data drawn from the same process.
→ https://en.wikipedia.org/wiki/Generalization_error

Graph of agents
A tree of LLM agents run by one coordinator that owns their statuses, parent links, and mailboxes. Agents create, message, wait for, and stop other agents through tools (Strix).
→ Agent memory and world-model repos

Group (algebra)
Set with associative binary operation, identity, and inverses.
→ https://en.wikipedia.org/wiki/Group_(mathematics)

Hash (cryptographic)
One-way compression function with collision resistance goals.
→ https://en.wikipedia.org/wiki/Cryptographic_hash_function

HCI / HMI
Human–computer / human–machine interaction.
→ https://en.wikipedia.org/wiki/Human%E2%80%93computer_interaction

Identity quality
Grade of an id: stable (an app-native id that survives screen changes), derived (computed from visible fields; cross-run grouping approximate), or ephemeral (valid in one parse run). From screenpipe’s IdentityQuality.
→ Agent memory and world-model repos

IEML
Information Economy MetaLanguage — constructed language with computable semantics (Lévy / INTLEKT).
→ https://en.wikipedia.org/wiki/IEML
→ snapshot: wiki-ieml.png

Impedance
Complex generalisation of resistance in AC circuits, $Z = R + jX$.
→ https://en.wikipedia.org/wiki/Electrical_impedance

ISA
Instruction set architecture — programmer-visible machine interface.
→ https://en.wikipedia.org/wiki/Instruction_set_architecture

---

K–O

Kinship
Socially recognised systems of relatedness.
→ https://en.wikipedia.org/wiki/Kinship

Language-tagged literal
A string with a BCP 47 language tag, "a mãe"@pt-BR. Tags compare case-insensitively. The tag is a claim about the text; nothing checks that the text is in that language.
→ IEML × RDF × Situation Monitor

Limit
Value approached by a sequence or function under a formal $\varepsilon$-$N$ / $\varepsilon$-$\delta$ definition.
→ https://en.wikipedia.org/wiki/Limit_(mathematics)

Loss function
Scalar objective minimised during learning.
→ https://en.wikipedia.org/wiki/Loss_function

Maxwell’s equations
Four PDEs governing classical electromagnetism.
→ https://en.wikipedia.org/wiki/Maxwell%27s_equations

Mental model
User’s internal predictive model of a system.
→ https://en.wikipedia.org/wiki/Mental_model

MinHash / LSH
MinHash estimates Jaccard similarity of two sets (here, 3-gram shingles of a name) from short signatures; locality-sensitive hashing buckets signatures in bands so only likely matches are compared. Graphiti uses 32 permutations, bands of 4, and accepts Jaccard ≥ 0.9.
→ Agent memory and world-model repos

Morpheme (IEML)
Basic semantic unit positioned in paradigms of opposition.
→ see IEML

Mutual information
$I(X;Y) = H(X) - H(X\mid Y)$ — shared information between variables.
→ https://en.wikipedia.org/wiki/Mutual_information

OntoLex
OntoLex-Lemon, the W3C community model for lexicons over ontologies: a LexicalEntry has written Forms and LexicalSenses whose reference is an ontology concept.
→ IEML × RDF × Situation Monitor

Ontology (information science)
Formal shared conceptualisation enabling inference and interoperability.
→ https://en.wikipedia.org/wiki/Ontology_(information_science)

Overfitting
Fitting noise or sample idiosyncrasies at the expense of generalization.
→ https://en.wikipedia.org/wiki/Overfitting

OWL 2 RL closure
All triples derivable from a graph under the OWL 2 RL rule set (W3C OWL 2 Profiles §4), computed by forward chaining. Comparing closures before and after a change shows what the change entails.
→ IEML × RDF × Situation Monitor

---

P–R

Paradigm (IEML)
System of semantic oppositions defining relative positions of morphemes.
→ IEML

POSIX
Portable Operating System Interface standards for Unix-like systems.
→ https://en.wikipedia.org/wiki/POSIX

Proof
Finite derivation of a statement from axioms and rules.
→ https://en.wikipedia.org/wiki/Mathematical_proof

Proof-carrying retrieval
An applied edge that carries a rule id and a certificate id on the payload a client receives.
→ Concepts governance

Provenance
Documented history of origin and transformations of an artefact or dataset.
→ https://en.wikipedia.org/wiki/Provenance

Psychophysics
Quantitative study of stimulus–sensation relationships.
→ https://en.wikipedia.org/wiki/Psychophysics

Public-key cryptography
Asymmetric key pairs enabling encryption or signatures without shared secrets.
→ https://en.wikipedia.org/wiki/Public-key_cryptography

RAG
Retrieval-augmented generation — LLMs conditioned on retrieved documents.
→ https://en.wikipedia.org/wiki/Prompt_engineering#Retrieval-augmented_generation

Ratio
Quotient relation $a:b$ or $a/b$ between magnitudes.
→ https://en.wikipedia.org/wiki/Ratio

rdf:type
The RDF property that makes a resource an instance of a class (a in Turtle). With RDFS or OWL inference, types can also be entailed, not only stated.
→ IEML × RDF × Situation Monitor

Reciprocal rank fusion
Merging several ranked lists by summing 1/(k + rank) per item. It is Graphiti’s default reranker for combining BM25, cosine, and graph-traversal hits (rrf, with k = 1 on a zero-based rank).
→ Agent memory and world-model repos

---

S–Z

Semantic Web
Stack of standards (URI, RDF, OWL, …) for machine-readable web data.
→ https://en.wikipedia.org/wiki/Semantic_Web

Set
Collection of distinct objects; foundation of modern mathematics.
→ https://en.wikipedia.org/wiki/Set_(mathematics)

SHACL shape
A W3C SHACL constraint set that names its targets (for example every sm:BriefSentence) and the conditions they must meet; a validator returns a report of violations, warnings, and infos instead of inferring new facts.
→ IEML × RDF × Situation Monitor

SKOS
Simple Knowledge Organization System (W3C): concepts, schemes, prefLabel/altLabel per language, broader/narrower, and exactMatch/closeMatch across schemes.
→ IEML × RDF × Situation Monitor

Sociomorphing
Perception of social capacities in non-humans via cues — not full anthropomorphism (Seibt et al.).
→ Sociotropia & sociomorphing

Sociotropy (sociotropia)
Personality orientation with high investment in interpersonal acceptance.
→ https://en.wikipedia.org/wiki/Sociotropy · PT https://pt.wikipedia.org/wiki/Sociotropia

Stoichiometry
Quantitative relations among reactants and products.
→ https://en.wikipedia.org/wiki/Stoichiometry

Swarm simulation
Many LLM agents with personas acting on a simulated platform, whose aggregate behaviour is read as a scenario outcome. MiroFish runs OASIS Twitter- and Reddit-like worlds; its output is a narrative report, not a calibrated probability.
→ Agent memory and world-model repos

Symmetry
Invariance of a structure under a group of transformations.
→ https://en.wikipedia.org/wiki/Symmetry

Tacit knowledge
Embodied skill hard to fully codify (Polanyi).
→ https://en.wikipedia.org/wiki/Tacit_knowledge

Threat model
Explicit statement of adversaries, assets, and capabilities.
→ https://en.wikipedia.org/wiki/Threat_model

Transformer (ML)
Neural architecture based on attention over token sequences.
→ https://en.wikipedia.org/wiki/Transformer_(deep_learning_architecture)

Triple term
An RDF 1.2 triple used as the object of another triple, <<( s p o )>>. It denotes a proposition and is not asserted; a reifier (r rdf:reifies <<( s p o )>>) carries metadata about it.
→ IEML × RDF × Situation Monitor

Turing machine
Abstract sequential computer defining computability.
→ https://en.wikipedia.org/wiki/Turing_machine

Uncanny valley
Dip in affinity for near-human artificial agents.
→ https://en.wikipedia.org/wiki/Uncanny_valley

Usability
Effectiveness, efficiency, satisfaction in context (ISO 9241).
→ https://en.wikipedia.org/wiki/Usability

USL
Uniform Semantic Locator — address of an IEML expression.
→ IEML

Zero trust
Security model that verifies continuously and assumes breach.
→ https://en.wikipedia.org/wiki/Zero_trust_security_model

---

Immutable and image-based operating systems

---

Snapshot gallery (core terms)

Offline snapshot omitted in this web edition. IEML · https://en.wikipedia.org/wiki/IEML · 2026-07-12

Additional snapshots are generated by the library’s curation scripts into .gitbook/assets/snapshots/wiki-*.png.

---

Keep this glossary short enough to scan; deep treatment lives in chapters.

Knowledge OS (LTRing / TEvEP / RUT)