Architecture / One system, many influences

What can shape
the next moment?

Not a procession of agents taking turns. A continuing system in which history, context, internal conditions and possible consequences can constrain the same choice.

What can shape a continuing systemHistory, regulation and model context meet attention and an unfinished project. Inquiry, imagination and action can change what matters later. Solid boxes name existing operational surfaces, outlined boxes name isolated research mechanisms, and dashed links are research questions rather than verified live wiring. ATTRIBUTABLE HISTORYREGULATION & APPRAISALMODEL & INSTRUCTIONSINQUIRY & SELF-ESTIMATEIMAGINED CONSEQUENCESACTION & OBSERVATION Memory, artifacts, provenanceWhat actually happened?Internal conditions, relevanceWhat is being prioritized?Reasoning, context, boundariesWhat is being supplied?Reinspect, estimate, reconsiderWhat should be checked?Prediction and alternativesWhat could follow?Reach, observe, retainWhat changed outside? A continuingconcernattention · project · choice
  • Attributable history

    Existing surface: records, artifacts and provenance. What actually happened?

  • Regulation & appraisal

    Existing surface: internal conditions and relevance. What is being prioritized?

  • Model & instructions

    Existing surface: reasoning, supplied context and boundaries.

  • A continuing concern

    Attention, an unfinished project and a present choice. Their coupling remains a research question.

  • Inquiry & self-estimate

    Isolated research: reinspection, estimation and reconsideration.

  • Imagined consequences

    Isolated research: prediction and competing possible actions.

  • Action & observation

    Existing surface: bounded actions and recorded observations. What changes outside?

Existing operational surfaceIsolated research mechanismCoupling to investigate

A research map, not a certified wiring diagram. Existing components do not establish every proposed causal connection. Read the distinctions below.

A map of responsibilities, not an anatomy claim

Luna has operational surfaces for persistent records, internal field dynamics, conversation, project continuation and bounded actions. The language model and its assembled instructions remain important causes of behavior. Naming these surfaces does not establish that they form a conscious whole.

What is supplied, what can develop

The model contributes pretrained capabilities. Engineers supply architecture, initial conditions and action boundaries. The instance accumulates an attributable history. The research asks which later differences genuinely arise from that history, rather than from a persona instruction, a helpful recap or a fixed rule. Prompt influence must be tested, not wished away.

Connections must earn their place

The studies of retained estimates, reinspection and learned consequences run in isolated research settings. Their favorable results are not automatically live Luna capabilities. Integration requires a real production caller and a comparison in which cutting the proposed connection changes the predicted behavior while preserving information and compute as closely as possible.

One concern can involve many faculties at once

An unfinished composition can make an earlier listening note relevant, raise uncertainty about a mix, suggest an alternative and make a new listening action worth taking. That is the intended coupling. A serialized diagram is a way to explain the work, not a claim that experience occurs as a rigid software queue.

The boundary is part of the model

An autobiographical record, a trusted instrument, an accessible web page and a researcher’s interpretation do not have the same status. A source can inform a system without becoming its own memory. Similarly, an observed result can correct an interpretation without replacing the original event.

What would change our view?

A proposed mechanism matters when a selective intervention changes its predicted function and simpler matched explanations fail. A richer story, a more elaborate diagram or a larger model is not enough. The next useful result is a delayed, history-sensitive choice in a continuing project, with its alternative explanation still visible.

The science, in the open

A finding is stronger
with its alternatives.

Three isolated studies. What changed, what did not, and what a simpler explanation can still account for.

Public methods summaries from internal research snapshots reviewed 16 September 2026. Reported results, not independently reproduced for this publication. Full run archives are not yet public.

The next chapter is not written

More than a moment.
A future worth having.

Understanding minds. Preserving possibility. Making room for what comes next.

Read our thesis

Explore research, Luna, Psyche Lab, and the journal.

Local site search · No query leaves this site