Axis Mundi
A Human Potential Simulator: an inhabitable world for discovery, construction, practice and return. Beyond the Grid is a narrative episode within it.
Research within the practice · Human–AI development
A human future,
explored with AI.
How can people and AI develop capabilities together—while people retain authority over the relationship?
One part of a wider practice in Reality Generation, world-building and human-capability design. These tracks investigate its mechanisms, experiences and consequences.
Part of the wider portfolio ←
Sovereign human–AI development Portfolio · October 2026
The question that connects it all
More capable.
Still able to choose.
An AI can become more useful while a person becomes less able to question it. A simulated world can offer new possibilities while quietly deciding what counts as progress.
My research begins with that tension. I explore how meaningful, continuing experience can expand what people understand and can do, while keeping assistance, interpretation and change open to human direction.
The programme explores sovereign human–AI development. Sovereignty here means practical authority: choosing a purpose, deciding when to delegate, challenging an interpretation, revising the relationship or leaving it. Learning matters; so do curiosity, enjoyment, appropriate assistance and the freedom to take another path.
Three projects · One connected inquiry
The experience. The context. The agency.
RealityOS
Operating environments for governed worlds: keeping context, identity and consequential state intelligible across different ways of engaging.
OCA
An architecture for inspectable, configurable and revisable agents: exposing selected structures of access, interpretation, learning and action.
Together, these projects propose a continuing person–agent–world relationship. Their individual implementations have different scopes; the complete developmental experience remains in progress.
Read the world · Locate the AI
An atlas of possibilities.
A relationship at its centre.
The maps hold the imaginative scope: an inner landscape, a world of lived experience and a wider field of possibility. The research asks how a person and an AI might participate in that landscape together.
The Cosmic Atlas
View full image ↗The whole proposition: a continuing relationship between an individual, an experienced world and wider possibilities. OCA provides the agent architecture; RealityOS provides operating context; Axis Mundi gives the journey a world.
The person
Chooses purposes, assistance and the terms of participation.
The AI / OCA
A modeled agent whose construction and selected changes can be examined.
The shared world
Axis Mundi supplies situations; RealityOS organizes context and continuity.
These are symbolic maps and design propositions. Their geography and terms such as “Conscious Field” are not measurements of human consciousness. The Inner Atlas’s RGE is an OCA ensemble; DU’s reality-generation service is a separate world-system dependency. Read the map’s vocabulary →
The programme
Three tracks.
Seven connected paths.
The tracks connect the purpose of the work, the mechanisms that make it inspectable, and the practice through which its claims are questioned.
IChosen human–AI
Purpose & relationship
Chosen human–AI
development
What changes through shared experience—and who gets to decide whether that change is valuable?
Axis Mundi treats adventure, construction, practice and return as parts of a developmental experience. Transformative TRUST asks whether the relationship expands a person’s possibilities on terms they can choose. Humane evaluation asks how reflection can support that inquiry without becoming a ranking of people.
- 01 Human potential through shared experience
- 02 Transformative TRUST
- 06 Humane evaluation
Start with the evidenceMara: construct, practise, return →Retained procedures change later computation →Reflection without a human scoreboard →
Open question Does a meaningful experience change what people can understand or do in a different context? Agent adaptation alone cannot answer this.
IIInspectable adaptive
Mechanisms & environments
Inspectable adaptive
systems and worlds
Can access, interpretation and consequential change be made understandable enough to question and revise?
OCA exposes selected structures of agent computation. RealityOS examines governed context, continuity and alternative views. Axis Mundi brings these questions into an inhabitable setting. Revision has several meanings here: changing a rule, flagging a dependent claim, revoking an operation or reopening a world.
- 03 Situated access & revisable cognition
- 04 Governed realities
- 05 Inspectable engagement
Start with the evidenceA rule changes after counterevidence →Recovery with conflicts and lost acknowledgements →Change the view, retain the context →Permission checked at the point of action →
Open question When does exposing a mechanism improve understanding—and when does it add effort or false confidence?
IIIAccountable
Evidence & method
Accountable
collaborative inquiry
How can human-directed, AI-assisted research retain its purpose while allowing evidence to change the plan?
The research process is itself an object of study. Development records let us reconstruct cases where work proceeded, stopped or was corrected. A useful result can be a boundary that held, a proxy that was relabelled or an ambitious interpretation that had to narrow.
- 07 Accountable human & AI research practice
- Connects every path in the programme
Start with the evidencePassing tests—and still stopping →Correcting a compelling representation →When more evaluation added no benefit →
Open question Do structured records help an independent reader reconstruct authority and unresolved claims better than ordinary development logs?
An instrument for the inquiry
See the mechanism
at work.
The learning workspace makes one comparison tangible: what can a retained procedure change in a later, bounded task?
The captured workspace shows eight encounters and four retained procedures. The linked research case documents a separate eleven-encounter sequence; the rule-refinement explainer below is another distinct record.
Read the broader learning evidence →
Opening detail of a local synthetic comparison showing retained procedures beside the no-memory condition. Retained October 2026 capture; it reports replay not yet checked in that session. These are computational results, not human learning measures.
Open full capture ↗A closer look · Recorded mechanism
What should change
when evidence changes?
This explainer follows an OCA rule-refinement record. It makes one small mechanism visible within the larger programme. It is a reconstruction of recorded stages, not a live agent run.
A useful shortcut, with a missing condition
Same event → inspect
The initial action rule requests inspection when accounts concern the same event. It does not yet require those accounts to be independent.
Retained rule
event=same_eventA dependent-account case exposes an unsupported inspection request.
Why this matters
A continuing relationship needs a way to be wrong and change. Here that idea becomes a retained rule revision. Whether a person understands, trusts or benefits from that revision is a separate research question.
What the work has taught us
Progress includes
finding the limits.
Agent learning is not human learning.
Retained procedures improved a bounded agent task. Human understanding, joint success and appropriate reliance still need separate observation.
The 11-encounter record →More machinery can add cost.
In three semantic-evaluation scenarios, outcomes were unchanged while the evaluator used extra reads. Inspection needs a purpose.
The equal-outcome result →A preserved state is only a beginning.
Recovery can retain identity and history without establishing meaningful continuity for a person—or completing a different system’s lifecycle.
Three distinct continuity cases →The wider intellectual landscape
Ideas before—and
beyond—the prototype.
The public essays trace questions about experience, human potential, qualitative reflection and the authority of measurement. Some extend into speculative field and consciousness accounts. They are authored propositions, with evidentiary requirements distinct from software demonstrations.
This portfolio draws those strands into a research programme; it does not claim to exhaust the entire development archive.
A research programme in motion
What becomes possible
when the relationship can evolve?
Explore the designs, examine the mechanisms, and keep the human question open.
Enter the evidence collection ↗The instruments behind the inquiry
See the systems
these questions come from.
Connectome construction and local execution, modeled access, RealityOS authoring and world inspection give the research concrete mechanisms to examine.
Powered by
RealityOS
AMIA
OCAThe connected explorer runs on RealityOS. AMIA supplies the wider platform context; OCA is presented through recorded demonstrations in this edition.