Update README.md

README.md

@@ -11,11 +11,37 @@ pinned: false
 short_description: ' A long-form conversational exchange instrument'
 license: mit
 ---
-This tool is discussed in the accompanying preprint: https://doi.org/10.5281/zenodo.18273459
+# Thread-Pulse
+Thread Pulse is a privacy preserving interaction analysis instrument for measuring stability, volatility, and regime transitions in long-form conversational exchanges.
+Rather than evaluating individual responses in insolation, Thread Pulse treats conversation as a temporal system. It analyzes turn-level behavioral signals, specifically response length and variance, across time to reveal how interaction dynamics evolve, stabilize, or fail under extended context. 
 
-# Welcome to Streamlit!
+Thread Pulse does not analyze semantic content, sentiment, or meaning. All measurements are derived from non-semantic structural properties of the interaction itself, making the instrument suitable for sensitive or private conversational data.
 
-Edit `/src/streamlit_app.py` to customize this app to your heart's desire. :heart:
+# What Thread-Pulse Measures
+Thread Pulse computes rolling statistics over conversational turns to identify:
 
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
+   1. Interaction volatility (response length variance)
+   2. Boundedness (whether variability remains constrained) 
+   3. Persistence (whether stability is sustained across time)
+   4. Time-to-Stability (the point as which a stable regime emerges, if at all)
+
+Stability is defined operationally as a period in which rolling variance remains below a configurable threshold for a minimum duration. This allows interaction regimes to be deteceted without assumptions about model internals, task correctness, or semantic coherence. 
+
+# What Thread-Pulse Is and Is Not
+Thread Pulse is an instrument, not a benchmark or evaluation score. It is designed to:
+
+   1. Observe interaction level behavior across scale
+   2. Detect regime transitions (exploratory - convergent - unstable)
+   3. Compare interaction dynamics across threads or models using a shared metric space
+
+It does not claim to measure: understanding, alignment, intelligence, intent or agency.
+
+# Current Scope
+This repository contains Thread Pulse v0.1, demonstrated on curated example datasets used to demonstrate the instrument. Thread Pulse operates on derived, non-semantic metrics only. Support for external dataset ingestion and streaming analysis is intentionally excluded from this release to preserve scope clarity and privacy boundaries. Thread Pulse is a part of a broader suite of interaction analysis instruments focused on understanding how stability forms and persist in Human-AI exchanges.
+
+Notes: This project was developed by Threadbourne. AI-assisted coding tools were used during implementation, consistent with modern software development practice. All conceptual framing, measurement logic, and evaluation criteria are human-directed.
+
+# Related Work
+An exploratory preprint describing temporal analysis methods and findngs presented here is available at:
+Beyond Content: Temporal Dynamics of Long-Form Human-LLM Interaction
+https://doi.org/10.5281/zenodo.18273459