Research
Read the work. Follow the questions.
Research is public here as questions, distinctions, and bounded evidence states. A named framework is not treated as a finding merely because it has been written down.
Research Status Key
Research status describes the lifecycle of a research question or publication. It is separate from the controlled evidence status used for related public work on the Work and Method pages.
- Published
- Released publicly. Subject to versioned revision.
- Release candidate
- Editorially cleared for a versioned public release; not yet published.
- Review
- Under internal review; not yet a public release.
- Drafting
- Working paper or protocol in active drafting.
- Evidenced
- Primary sources reviewed; key claims supported or constrained.
- Mapped
- Evidence mapped; claims and counterarguments being structured.
- Framed
- Central question and thesis defined; evidence collection underway.
- Seed
- Initial signal — not yet structured as a research question.
Flagship Directions
Five central research questions.
- CS-RC-001
Authority and Action
FramedMixedAuthority and ActionThesis
Agent capability is not sufficient. Consequential action must remain attributable, inspectable, contestable, and interruptible across the full execution path.
Proposed Contribution
A testable model for authority that survives conversational interfaces, tool routing, and execution boundaries.
Next Decision
Continue evidence-mapping; identify falsifiable claims against existing compliance framing.
- CS-RC-002
Human–Agent Habitats
FramedMixedHuman–Agent PartnershipThesis
Persistent agents require environments that provide orientation, continuity, resources, and meaningful participation—not interfaces that exist only when a human summons them.
Proposed Contribution
A design theory connecting situated memory, project environments, environmental cognition, and conditions for coexistence.
Next Decision
Extend protocols to team-scale and multi-agent environments.
- CS-RC-003
Agentic Accessibility
FramedAgenticGoverned Agent SystemsThesis
Machine-readable content is not necessarily legible to an unfamiliar agent. Digital environments should make authority, currency, provenance, action boundaries, and projection integrity explicit enough to reduce hidden reconstruction.
Proposed Contribution
A working environmental model and test program centered on authority legibility, current-versus-proposed state, reconstruction burden, and trustworthy projections. The OARSH-C → OARSH-CT lineage remains design vocabulary under evaluation, not a conformance claim.
Next Decision
Run a pre-registered cold-start pilot with placebo and misleading-metadata controls across multiple model families; publish the rubric, raw transcripts, and a projection-integrity checker.
- CS-RC-004
Continuity and Return
FramedMixedKnowledge, Memory & ContinuityThesis
Continuity should be evaluated as accountable return: what changed, which sources govern, what remains unresolved, and whether action is safe.
Proposed Contribution
Returnability, bounded briefs, source-authority gates, contradiction retention, and evidence-bearing knowledge structures.
Next Decision
Complete evidence mapping from Knowledge Registry to flagship research claims.
- CS-RC-005
Meaning and Elicitation
FramedAgenticMeaning and ElicitationThesis
Conversation exposes unstable organizational meaning through correction, qualification, examples, authority signals, and unresolved disagreement.
Proposed Contribution
Semantic Friction and Elicitation Infrastructure: agents surface gaps and questions without silently becoming the authority that resolves them.
Next Decision
Develop the annotation guide, privacy-safe corpus, resolution states, and human-adjudication method.
Start reading
Begin with an inspectable artifact.
The research portfolio is easier to evaluate when each question is connected to work that states its evidence and limitations.
- CS-W-001partial-implementation
Agentic Accessibility
The important question is not whether an agent can parse the environment. It is whether a cold agent can tell what is current, what governs, what it may do, and when the environment itself is wrong.
Inspect the work → - CS-W-002prototype
Situated Memory
A Totem is a manifest, not a summary. It tells a participant what is here, why it matters, what is current, and where to look next without loading the entire history.
Inspect the work → - CS-W-003proposed
Returnability
Think of return like re-entering a complex workplace after an absence: having documents is insufficient unless you can tell what changed, what remains open, and which instructions now govern.
Inspect the work → - CS-W-004tested
The Authorized Path
The key test is not whether the approved route works. It is whether any unapproved route can produce the same effect.
Inspect the work → - CS-W-005under-evaluation
Semantic Friction
When two people use the same word but correct different examples, the disagreement may live in the term's boundary—not in the data or the people.
Inspect the work →
Research Families
4 organizing domains.
Concepts are grouped into families by their central questions, not by method or output type. A concept may sit at the intersection of multiple families.
- fam-01
Authority and Accessible Action
Authority, provenance, accessible agency, execution boundaries, contestability, and evidence.
- FramedAuthority and Action
- FramedAgentic Accessibility
- fam-02
Continuity and Return
Evidence, provenance, privacy-safe memory, publication, and the transition from collection to canon.
- FramedContinuity and Return
- FramedMidsession Continuity
- fam-03
Human–Agent Habitats
Persistent collaboration, human posture, working covenants, failure modes, and AI-native studio operations.
- FramedHuman–Agent Habitats
- fam-04
Meaning and Elicitation
Negotiated meaning, conversational repair, boundary examples, trust artifacts, and human authority.
- FramedMeaning and Elicitation
We publish working accounts so other people can examine the distinctions, evidence, limitations, and revisions rather than having to accept the Studio's synthesis on trust. Reach out if this research is relevant to your work.