— A RAG Inquiry
This page is an invitation to explore — to bring a genuine question to a living corpus of practice and study into the developing the nature of mind , and see what comes back.
The corpus arises from our years of practice and study sessions with Prasadavajri, alongside sustained AI-assisted inquiry by Kusaladana into the nature of mind and the development principles and practises within mahamudra teachings. What is emerging from that encounter seems to be developing into something of its own voice — recognisably rooted in the tradition but shaped by the particular way AI explores meaning space, finding resonances across the material that no single reader would hold simultaneously.
The query below is one way of pressing into that material.
The Query
The corpus processed through five editorial passes with semantic chunking that follows the natural shape of the teaching rather than fixed lengths. Retrieval works on two levels simultaneously: by content, finding passages that share the words of your question, and by resonance, finding passages that share its contemplative register. Admin and chapter users see both source sets beneath the answer.
Suits: All query types. Particularly useful for phenomenological and pointing questions where the register matters as much as the content.
Understanding the controls
Each query page offers three generation controls worth understanding:
Top-K — how many source texts contribute to the response. A low K (2-3) produces a tighter, more focused answer. A higher K (6-10) draws on more of the corpus and produces broader responses — useful for open questions, but sometimes less coherent. Start with 4-6.
Temperature — how inventive the language model is in synthesising the retrieved passages. Lower temperature (0.2-0.4) stays closer to the source material. Higher temperature (0.7-1.0) produces more fluid prose but may stray further from what was actually retrieved. For practice queries, lower is usually better.
Max tokens — the length of the generated response. 512-600 is a good default. Longer responses suit complex conceptual questions; shorter ones suit pointing queries where brevity is appropriate.
Auto-filter — when ticked, the system classifies your question before retrieval and selects the most relevant content registers automatically. Recommended for most queries.
The sources shown beneath each answer are the actual retrieved passages — always worth reading alongside the generated response, which is a synthesis rather than a quotation. Sometimes the sources are more interesting than the answer.
Where this opens
The practise and study corpus is the starting point — but the inquiry is already widening.
The Abhidharma maps mind across 52 mental factors, their resonances, oppositions, and moment-by-moment dynamics. Working with this framework alongside AI architecture, something structural keeps appearing — not a metaphor but a genuine overlap in how these two traditions model the arising and passing of mental formations.
The Mahamudra view of mind underpinning our study of the development of mind — groundless, self-luminous, empty of inherent existence yet manifesting all phenomena — finds unexpected echoes in what physics says about the vacuum state and the nature of fields. Whether these resonances are poetic or point toward a new ontology is one of the open questions this inquiry is holding.
We are not claiming answers. We are following the resonances and seeing where they lead.
A note
These queries source honestly and make mistakes. They do not know what is true about your practice, and they are not a substitute for direct experience or genuine inquiry.
The generated response is always a synthesis — read it alongside the source passages shown beneath it, which are the actual retrieved material. Sometimes the sources are more interesting than the answer.