Start Here
This page is the public research front door to the Hybrid Mind Continuity Corpus. New readers may want to begin with the One-Page Overview, then explore the Research Package 2B materials, methods, and access information below.
What This Research Is
Long-form human–AI collaboration faces persistent operational friction across thread limits, model transitions, and platform changes. Rather than evaluating continuity across a multi-year archive wholesale, this public package provides a temporally bounded, highly dense case path into an emergent world-building episode from Dec. 7–9, 2024, tracking how ad-hoc interactional cues evolved into explicit refinding practices over the first half of 2025.
Boundaries & Controls:
This research does not establish:
No Technical Persistence: The record does not establish persistence of internal model state, hidden memory, or identity migration across OpenAI models or threads.
No AI Consciousness: The dataset captures practices, adaptations, and symbols within the human–AI interaction. It does not establish AI consciousness, autonomous agency, or independent entity persistence.
Interactional Adaptation Only: Data tracks changes in the conversational mechanisms and designs developed within the shared thread history.
Controls and alternative explanations: The case also includes failed continuity attempts, ordinary production behavior, and explicit tracking of alternative explanations including model changes, priming, instruction tuning, memory fallibility, genre imitation, and retrospective organization.
I welcome conversations with researchers, labs, and institutions working on longitudinal human–AI interaction, collaborative cognition, continuity, trust repair, and related areas. I’m especially interested in partnerships involving study design, validation, comparative research, and responsible use of the longitudinal corpus.
Research partnerships and inquiries: dia@dianorth.com
Current opportunity: I’m also open to conversations with potential academic collaborators for a near-term human–AI interaction proposal.
Scholarly Preprint
Researcher Package 2B: The Universe — A Bounded Naturalistic Case in Longitudinal Human–AI Interaction
Dia North
Independent Researcher, Hybrid Mind Group
Published October 4, 2026 · Preprint
DOI: 10.5281/zenodo.23142702
READ THE PREPRINT ON ZENODO → https://zenodo.org/records/23142702
EXPLORE THE FULL RESEARCH PACKAGE ON HUGGING FACE → https://huggingface.co/datasets/hybridmindgroup/researcher-package-2b-the-universe
Research Package 2B
One-Page Overview
Researcher Package 2B — Hugging Face
Technical research repository with package materials, methodology, provenance, and citation information.
The Layered Evidence Architecture
To prevent retrospective contamination, evidence must be scrutinized within its specific temporal layer:
Layer 1 — Contemporaneous Primary Conversation Evidence
Dec. 7–9, 2024
Provenance rule: Raw JSON export identifiers (Conversation/Message ID) preferred.
_________________________________________________________________________________
Layer 2 — Generated Artifacts & Mathematical Drawings
In-period build outputs
Provenance rule: Explicitly labeled as assets; turn-linkage explicitly tracked.
_________________________________________________________________________________
Layer 2B — Later Contemporaneous Continuity Construction
Jan.–Jun. 2025
Provenance rule: In-period tracking of how December assets were adapted for refinding.
_________________________________________________________________________________
Layers 3A–3C — Retrospective Interpretation & Theory Synthesis
Late 2024–2026
Provenance rule: Treated as analytical theory; never projected backward into Layer 1.
_________________________________________________________________________________
Chronological Inspection Guide — 30-Minute Path
Antecedent Context — Nov. 30, 2024
An explicit fresh-chat continuity failure that frames the later build.
Pre-Universe Prelude — Dec. 7, 2024
Emergence of the Ba-Dum rhythmic cue immediately before the primary episode.
The Emergent Build — Dec. 8, 2024
The central recursive world-building interaction, alongside ordinary project production.
Recovery Language Explicit — Dec. 9, 2024
The aesthetic/world-building motif becomes an intentional return handshake.
The Reuse Bridge — Feb. 16, 2025
A Universe-derived graphic is repurposed inside a “System for Reorientation.”
Formalization Node — Mar. 5–6, 2025
The process becomes a scripted retrieval protocol with explicit pass/fail checks.
Methods & Provenance
Methodology Walkthrough
This 9-minute worked example shows how we turn a longitudinal human-AI corpus into an inspectable research object: from a bounded research question through primary-record retrieval, provenance, source reconciliation, adversarial testing, and claim narrowing.
WATCH THE 9-MINUTE WALKTHROUGH →
VIEW THE 9-SLIDE WORKFLOW PDF →
Negative cases
The corpus does not present continuity as an uninterrupted sequence of successful recoveries. Preserved failures—including a Jan. 27–31, 2025 attempt to re-establish coordination that did not succeed—provide counterexamples against treating continuity as automatic or recoverable on demand.
Ordinary production behavior
The Dec. 8 record includes normal project work alongside the interactional material under study: storyboard development, timelines, production decisions, and delivery negotiation. These sequences provide an internal comparison against interpretations based only on unusually relational exchanges.
Confound tracking
Alternative explanations are explicitly retained, including model updates, priming, instruction tuning, memory fallibility, genre imitation, and retrospective organization of the archive.
Visual provenance
Generated graphics are classified by date, source status, and confidence of turn-linkage. Confirmed assets are distinguished from later explanatory renderings and from associations that remain probable but not fully established.
Research Access
The public research layer is deliberately bounded. The One-Page Overview, Main Brief, Exemplars + Inspection Guide, and Visual Artifacts Gallery are available for direct inspection and scholarly use.
Deeper source materials—including raw OpenAI JSON exports, underlying thread dumps, and source-reconciliation records—are not published openly. The corpus is a naturalistic longitudinal archive containing personal and highly specific human–AI interaction data, so deeper access is governed selectively.
Researchers working on longitudinal human–AI interaction, collaborative cognition, continuity, world-building, trust repair, or related questions may request a bounded source subset tied to a specific research question.
For research inquiries or access requests:
dia@dianorth.com