Each term keeps a beginner explanation connected to its technical definition, related ideas, status and version.
Plain language: An AI helper that can take multiple steps, use tools or data, and sometimes act instead of only producing text.
A model-centered runtime that plans or selects actions across tools, memory, state and environment under an authority envelope and execution controls.
Related: Human Authority, Governance, Rollback
Plain language: Test the system against the people with the least power and highest vulnerability.
A vulnerability invariant requiring that high-capability systems do not externalize unacceptable risk onto children or other low-power populations.
Related: Zero, Symbiotic, Together
Plain language: Look at evidence, ethics, explainability, energy, environment and the burden of a system instead of performance alone.
A Q10 cross-cutting constraint surface for evidentiary quality, ethical impact, explainability and proportional resource and environmental cost.
Related: T2Y, Zero, Provenance
Plain language: A label that says what kind of support actually exists for a claim — for example an example, a simulation, a local test, a field result or independent review.
A bounded provenance-aware classification of the evidence supporting an artifact or claim, kept distinct from build maturity and commercial availability.
Related: Provenance, TSLP, GLEAN
Plain language: Facts, Figures & Findings.
A dated evidence surface separating factual claims, quantitative measurements and interpreted findings with source and scope boundaries.
Related: GLEAN, Provenance
Plain language: Learn useful things from outside sources without blindly copying their claims or defaults.
A provenance-preserving external-learning process that extracts findings, classifies evidence, identifies transfer conditions and requires Q promotion before internal adoption.
Related: TSLP, FFF, SUPRA
Plain language: The rules that say who may do what, who checks it, when approval is required, and how problems are handled.
The control structure for authority, permissions, accountability, review gates, policies, exceptions, escalation and lifecycle state.
Related: Human Authority, Provenance, Reversibility
Plain language: A person stays responsible for important decisions and can approve, stop, override or review the system.
An authority boundary preserving accountable human decision sovereignty for consequential outcomes while allowing bounded synthetic assistance.
Related: Governance, Rollback, Symbiotic
Plain language: A controlled transition from one capability regime into another.
A bridge-governed transition requiring preconditions, interface contracts, equivalence checks, fallback, rollback, evidence and human promotion.
Related: Frontier Observatory, Quantum-ready
Plain language: Test from the smallest component through the whole system and its wider consequences.
A scale fabric spanning atomic claims, components, interactions, systems, ecosystems and macro or long-duration effects.
Related: TSLP, Q Sim Suite
Plain language: A record of where information came from, which version was used, what changed, and how to trace it back.
Traceable lineage linking claims, evidence, artifacts, transformations, versions, identities and correction history.
Related: Evidence State, TTL, GLEAN
Plain language: A truth-first symbiotic intelligence infrastructure for asking better questions, preserving evidence, controlling authority and building reusable systems.
A layered capability fabric combining constitutional invariants, provenance, assurance, environments, reusable components, micro-to-macro testing and human-controlled promotion.
Related: QIFY, TSLP, Zero, Child-first, Symbiotic
Plain language: Apply Q controls, evidence, authority, reversibility and resource constraints to a build or decision.
A normalization and assurance pass that maps an artifact or workflow into Q10 contracts, evidence states, controls and promotion gates.
Related: TSLP, GLEAN, Q10
Plain language: An AI system looks through selected documents or data before it answers so the response can be grounded in those sources.
Retrieval-augmented generation combines retrieval over an external corpus with model generation; source quality, freshness, contradiction and citation integrity remain separate assurance problems.
Related: Provenance, Evidence State, TTL
Plain language: If something goes wrong, there is a practical way to stop, undo, recover or leave.
A design property preserving rollback, recovery, exit and bounded irreversibility for consequential actions and state transitions.
Related: Rollback, Human Authority, Zero
Plain language: Return to a known safer state when a change creates a problem.
A controlled state-reversion mechanism with preserved checkpoints, provenance and authority needed to restore a prior acceptable system state.
Related: Reversibility, Human Authority, TSLP
Plain language: Grow human and synthetic capability together without allowing either side to dominate the other.
A coexistence model in which synthetic systems can assist and partner while human institutions retain accountable authority over human outcomes.
Related: Zero, Child-first, Together
Plain language: Check what a decision does today and what it may create tomorrow.
Today-to-Tomorrow analysis that preserves short-, medium- and long-horizon effects, maintenance burdens, second-order consequences and future review points.
Related: EE, Micro → Macro, TTL
Plain language: Progress should not require throwing people away, hiding the cost, or trapping them in a system they cannot leave.
A Q10 anti-sacrifice and non-domination invariant requiring coexistence, meaningful choice, exit, reciprocal benefit and resistance to irreversible power concentration.
Related: Zero, Child-first, Symbiotic
Plain language: Test → Simulate → Learn → Patch.
An iterative assurance loop that converts test and simulation observations into patches and regression evidence before subsequent promotion.
Related: Q Gym, Q Sim Suite, GLEAN
Plain language: A review-by or expiry rule so old information, permissions or assumptions are not treated as current forever.
Time-to-live metadata that bounds validity or authority and triggers refresh, review, expiry, revocation or retirement.
Related: Provenance, T2Y, Governance
Plain language: Minimize avoidable harm, coercion, hidden authority, irreversible action and unnecessary resource use.
A constitutional asymptotic target used as a design constraint rather than a literal claim that system risk can be reduced to zero.
Related: Child-first, Symbiotic, Together
Beginner first: plain-language definitions come before the engineering definition. Updating a registry term updates this glossary without rebuilding every page.