Founder / operator
Garth Noel.
I work on AI systems, software architecture, autonomous work, and the authority problem underneath increasingly capable machine systems.
I build before I explain.
I design, fund, test, and operate the work directly. AI models and coding agents are development instruments in that process; they are not decision authorities.
My current focus includes persistent agentic systems: software that can perceive changing environments, reason across longer horizons, coordinate work, interact with other machine systems, and participate in emerging machine-to-machine economies while remaining inspectable and bounded by human authority.
The recurring question is not whether machines can become more capable. It is how capability can increase without silently inheriting authority over execution, resources, communication, evidence, or the operator.
The programme behind much of the work.
ANANS is the active research-and-build programme through which I explore these questions in software. It should not be reduced to an AI governance tool, compliance product, agent framework, DevSecOps system, or consulting method. Those are comparison points around parts of the stack, not definitions of the whole.
ANANS remains privately developed. Public evidence is deliberately narrower than the private system. Where a claim is public, the preferred form is an inspectable artifact, a bounded demonstration, a recorded failure, or a reproducible receipt.
What this publication is not.
- Not an ANANS mirror.
Private runtime, trade-secret material, confidential research, and operational state do not belong here. - Not an authority channel.
A published idea, model output, essay, or research note cannot authorize execution inside ANANS. - Not a product-readiness claim.
Experiments and synthetic demonstrations remain experiments and synthetic demonstrations unless stronger evidence exists. - Not a content machine.
Nothing is automatically promoted from research output to publication.