The line AI
doesn’t cross.
PulZ is a deterministic governance kernel — the guardrails that decide whether an AI action is allowed, based on evidence and confidence, and keep a tamper-proof record of every call. It’s what makes owning your AI safe.
Evidence
Decisions are backed by what’s actually known — not a model’s gut feel. If the evidence isn’t there, the action doesn’t pass.
Confidence
Actions are gated by confidence. Low-certainty moves get held, escalated, or refused instead of waved through.
Audit
Every decision lands in a tamper-proof chain. You can prove, later, exactly what was allowed and on what basis.
Guardrails shouldn’t guess.
You can’t use a model to police a model — it drifts and hallucinates like any other. PulZ is deliberately small, deterministic, and zero-dependency: the same inputs always reach the same verdict. That’s what lets you trust it as the layer between an AI’s intent and the real world.
What is PulZ?
A deterministic governance kernel for AI. It’s the layer that decides whether an AI action is allowed to happen — based on evidence and confidence — and writes a tamper-proof record of what was decided and why.
What does “deterministic” buy me?
Predictability. The same inputs always produce the same governance decision, so the guardrails don’t drift or hallucinate. You can reason about what the system will and won’t allow — which is exactly what you want from a safety layer.
Is PulZ the same as ORA?
No. They’re two separate products. ORA is the orchestration gateway that runs and coordinates AI on your hardware. PulZ is the guardrails — the kernel that gates actions and keeps the audit trail. They complement each other but are not the same thing.
What is it built on?
A focused, zero-dependency TypeScript kernel — small enough to trust, with confidence thresholds on actions and audit chains you can’t quietly rewrite. It’s in active development.
Need AI you can trust to behave?
PulZ is in active development. Email info@3d3d.ca to talk about the guardrails layer for your AI.