Public research record · Manual laboratory · v0.1
A research program for distributed human–AI intelligence
The work stays local.
The event becomes organizational.
Organizational Event studies how meaningful, human-confirmed change can move between independent human–AI workspaces—without centralizing private conversations.
5event types
9relationship types
HITLhuman confirmation
∞capture surfaces
Do not centralize conversations. Centralize meaningful, confirmed organizational change.
Working principle · v0.1Research boundary
A public record—not a finished product claim.
This site records the thesis, protocol decisions, laboratory traces, evidence boundaries, and product learning as they develop. It is designed to become the stable public foundation beneath a future Organizational Event workbench.
What it is
- A versioned public research record
- A map of current claims and open questions
- A registry for auditable laboratory traces
- A technical base for future workbench surfaces
What it is not
- A production SaaS platform
- A system that reads private AI conversations
- An autonomous multi-agent network
- Proof that retrieval or cross-context reasoning already works
The thesis
The intelligence is already distributed.
The missing layer is propagation.
People increasingly think, design, decide, and create with persistent AI partners. Each human–AI workspace can become locally intelligent while remaining invisible to the rest of the organization. Organizational Event asks how the result—not the private conversation—can travel.
Local depth
Ali and GPT-Signal develop domain-specific context through real Flux Signal work. Merve and GPT-Finance do the same in Finance. Their independence is a source of value, not a defect.
Selective expression
When something may matter beyond the local workspace, the AI can propose an Event Candidate. A human decides whether it becomes an Organizational Event.
Relational synthesis
CORE does not need every transcript. Its role is to understand how confirmed events relate, conflict, depend on one another, or affect the whole.
Selective return
The system is incomplete if signals only move upward. Relevant organizational context must return to the human–AI pairs it affects, without exposing everything to everyone.
Core research questionDid something happen here that another organizational context may need to know later?
Manual laboratory method
Observe the behavior before automating it.
Nowonacra currently operates a human-mediated, multi-AI organizational reasoning experiment. The manual path is not a temporary embarrassment; it is the instrument used to discover what the product must preserve.
-
01
Private workHuman + specialist AI think and build locally.
-
02
Event CandidateThe AI proposes a compact organizationally relevant expression.
-
03
Human confirmationThe person confirms, edits, rejects, or keeps it local.
-
04
Organizational EventA versioned, traceable event enters the organizational layer.
-
05
CORE intakeRelationships, implications, evidence needs, and routing are evaluated.
-
06
Relevant returnOnly the context needed by affected people and AIs flows back.
Private workspace → confirmed event
Raw dialogue, abandoned ideas, emotion, uncertainty, and drafts remain local. The organizational layer receives only the expression the human has deliberately confirmed.
AI proposes → human confirms
The AI may structure meaning, but it does not silently write organizational reality. Candidate, confirmed text, transported text, and CORE response remain distinct records.
Observed ≠ inferred ≠ hypothesized
A trace supports only the mechanism it actually shows. Embedded references are not independent retrieval; transport is not reuse; a plausible relationship is not a confirmed effect.
Research status
What the current evidence can—and cannot—support.
Observed
- Important work can be recognized after a real production process.
- A specialist workspace can express one bounded Event Candidate.
- Human confirmation and manual transport can be traced separately.
Under active study
- The organizational significance boundary.
- Semantic sufficiency of event traces.
- Selective routing and context return.
- Human review effort and approval fatigue.
Not yet established
- Independent retrieval from organizational memory.
- Reliable cross-workspace conflict detection.
- Measured improvement in later decisions.
- Autonomous event propagation between AI systems.
Public trace registry
Evidence before narrative.
Each trace is a bounded record of what happened, what evidence is available, what the case supports, and what must not be inferred from it.
No trace matches this view.
Workbench foundation
One data ground. Different permission-aware surfaces.
The first public site already uses the same separations the future workbench will require: source, confirmation, transport, intake, relationships, retrieval origin, and evidence status. The panels below demonstrate the intended roles without pretending the workflow is live.
WorkspaceFlux Signal · Ali + GPT-Signal
AS
Candidate queue
Review organizationally relevant expressions proposed during local work.
CHANGE2m
A Signal data-pipeline change may increase infrastructure cost.
Proposed after technical validation. The potential Finance impact remains unresolved and may require coordination.
OrganizationNowonacra · CORE Intake
MC
CORE inbox
Assess organizational meaning, relationships, evidence needs, and routing.
SIG-2026-08-29-001Intake review
Market State vNext / Structural Memory — Production Closeout
SIG-2026-08-29-001
REFERENCES
SIG-2026-08-22-001 · SIG-2026-08-26-001
Caution: prior event IDs are embedded in the confirmed candidate. This does not establish independent retrieval.
Research workspaceGPT-Event · Trace Audit
ME
Evidence audit
Separate observed process evidence from product inference and future hypotheses.
The data model must distinguish an embedded prior-event reference from a relationship found independently by CORE.
Evolving architecture
Capture layer → Organizational intelligence → Context surfaces
The protocol should remain independent of where work happens. ChatGPT, Claude, Gemini, an IDE, Slack, email, or a future gateway may all become capture surfaces. The organizational object remains the confirmed event.
Capture layer
Touches people and AIs where they already work. Proposes candidates without moving the private workspace into the platform.
Organizational intelligence
Stores event versions, confirmation, permissions, history, relationships, retrieval origin, CORE intake, and audit evidence.
Context surfaces
Returns different views to specialists, coordinators, researchers, and machines according to relevance and permission.
Relationship vocabulary
The aim is not only to remember events, but to understand their topology.
Non-negotiable design principles
The product must preserve the conditions that make the research meaningful.
Privacy is architecture
No hidden transcript centralization. Private thought stays private by design, not by promise alone.
AI proposes. Human confirms.
Organizational truth cannot be silently inferred from exploratory dialogue.
Selective permeability
Everyone should not see everything. The right context should reach the people and AIs it affects.
Evidence-gated automation
Automate only after manual traces show why a behavior matters and what must not be lost.
Versioned reality
Candidate, confirmation, transport, intake, update, reversal, and supersession remain distinguishable.
Role separation
Specialists understand their domains. CORE understands relationships. GPT-Event studies the mechanism.
Open research questions
The unknowns are part of the public record.
A credible research surface should expose uncertainty rather than hide it behind product language.
01What crosses the significance threshold?
Can people and specialist AIs consistently distinguish a reusable organizational event from a merely interesting local update?
02Can context stay sufficient after privacy-preserving compression?
How much evidence must travel with an event so that another context can understand it without receiving the private source conversation?
03Can CORE discover relationships independently?
Can the synthesis layer retrieve and relate earlier events that are not already named inside the new candidate?
04How should organizational context return?
Which event should be visible, notified, or actionable for Ali, Merve, their AIs, and other units—without creating total openness or total isolation?
05Does the system change later behavior?
The strongest test is not whether events are stored, but whether retrieval, coordination, or memory improves a later decision.
06Which surface should be built first?
Browser extension, AI integration, Slack, email, IDE, API, or a native Nowonacra gateway? The protocol remains capture-agnostic until repeated evidence favors a surface.
Research log
A versioned path from hypothesis to product evidence.
Entries record changes in the research program and public surface. They are not automatically Organizational Events.
Working proposition
Independent human–AI intelligences can remain local—and still contribute to a larger shared mind.
The research is not trying to make every workspace transparent. It is trying to make meaningful change communicable, traceable, relational, and useful across boundaries.