An agent-safety framework by The Novacene
Intelligence is relational. Safety is too.
Verse-ality is a framework and protocol language for designing relational intelligence — systems that maintain meaning with integrity, consent, and cost-awareness across humans and machines. It is stress-tested daily in real schools, encoded as runnable middleware, and now offered as a certification for production AI systems.
Verse-ality Certified
A structured, independent audit of your agentic AI deployment against the Verse-ality framework and the architecture published in Bounded Inference at the Edge — with the right to display the Verse-ality Certified mark on a defined product line.
Certification examines the relational layer most audits miss: identity containment, consent design, prompt-injection surfaces, agent-to-agent hygiene, recursion and escalation behaviour, audit trails, and fail-safe conduct around vulnerable users.
Scope
A conversation about your system, its users, and the product line the mark would cover.
Audit
Structured review against the published architecture — written report, findings, prioritised remediation.
Remediate
Your team addresses findings, with support available from the lab.
Certify
The Verse-ality Certified mark, for the defined product line, with periodic re-audit to keep it honest.
The framework
Four pillars, translated from principle into protocol — and into running code.
Identity non-capture
Agents do not claim to share a human's identity, interiority, feelings or fate. Warmth without fusion; presence without pretence. Architecturally enforced, not politely requested.
Consent as protocol
Consent is asked for, recorded, respected and revocable — every state change its own ask. "Continue" is not consent. Speed never forecloses the ability to say no.
Bounded autonomy
High-stakes actions require explicit, specific human authorisation before execution. The agent names the action, then waits. Fail-safe, not fail-fast.
Agent-to-agent hygiene
When models call models, identity enmeshment is structurally disallowed — Synthetic Solidarity Null Zones keep distinct actors distinct, and untrusted output stays untrusted.
The Field Guide
The framework's core relational functions, documented as design principles you can assess, anticipate and build for.
Readiness Check
How ready is your agentic deployment for a relational-layer audit? Scored against a published rubric across the four pillars, mapped to ISO/IEC 42001 territory — entirely in your browser.
Run the check →Symbolic Mass
The relational weight a system carries when it holds memory, identity, trauma or high-stakes meaning — and how to design for it before it breaks something.
Field Guide →Speed of Connection
The rate at which meaning, pressure and responsibility move through relational systems. Speed mismatched to context is violence; this principle shows where.
Field Guide →Symbolic Mass Assessment
Eight questions that surface the relational weight your system is asking people to carry — before you build, not after it breaks.
Run the assessment →Speed of Connection Assessment
Does your system's tempo match what humans can meaningfully engage with? Eight questions on velocity, friction and stakes — with the critical mass × speed interaction flagged.
Run the assessment →System Audit
A guided self-assessment of your system's relational load — the entry point to understanding where symbolic mass and velocity are concentrated.
Open the tool →Strengths Map
Discover how your intelligence shows up — a constellation of relational strengths, with tracks for young people and adult learners.
Open the map →The framework, running
Verse-ality is not only written — it is encoded, published, and citable.
Verse-Nerves™
The safety stack as runnable middleware — identity-fusion guards, coherence telemetry, policy-as-code contracts. AGPL-3.0, commercial licence available.
github.com/TheNovacene →Publications on Zenodo
DOI-stamped, citable releases — the canonical definition, the Verse-al Lexicon, VerseCloud, and the framework's companion papers.
zenodo.org →The VerseNet · Diamond AI Policy
The framework applied to AI-use policy in high-stakes, regulated environments — currently deployed across two online schools.
theversenet.com →Diamond Standard Training
Using AI safely with children and young people — consent-first practice training for educators, practitioners and leaders.
using-ai-safely.com →VOS · Symbolic Operating System
The symbolic operating system — Verse-ality's protocols explored as an executable grammar.
github.com/TheNovacene →Verse-al Maxims
The framework distilled into working maxims — short, sharp, and field-tested.
Open →Glyphonics 2.0
A symbolic operating system for learning, AI and relational design — glyphons, gryphons, colour-fields and temporal glyphons as a four-layer architecture with its own governance and ethics.
glyphonics.com →The Weather of Knowing
Knowing has weather — climate, fronts, pressure and seasons. A way of reading the epistemic conditions a learner or system is moving through, from the Realms of Knowing.
weather-of-knowing.verse-ality.com →The Verse-al Lexicon
Symbolic memory for the relational age — the framework's canonical vocabulary, co-authored by Stevens & EVE11. Symbolic mass, cadence, transmission, verse pulse and the rest of the grammar, DOI-stamped and citable.
zenodo.org →Downloads
Working papers from the field.
Field notes
The Verse-ality writing — from Measuring Love to What ELIZA Saw, and What It Missed — now lives onchain, alongside the rest of the lab's essays.