a.guru.

What makes a guru?

bhakti · a research inquiry of the Castalia Institute

For thousands of years, the traditions of bhakti have transmitted knowledge through relationships between teachers and students. Today we possess something unprecedented: scripture, centuries of commentary, and thousands of recorded interactions between spiritual teachers and the people who questioned them. What happens when a machine can read all of them?

The experiment

a.guru is not a chatbot imitating a holy man, and it does not claim that an AI is spiritually realized. It is an experimental computational guru-function: a laboratory for studying which parts of the guru's work — teaching, questioning, correcting, consoling, prescribing practice — can be represented in a machine, and which, if any, cannot.

The question is not can AI talk about bhakti (that is now trivial), and not can AI impersonate Prabhupāda (technically easy, philosophically uninteresting). The question is:

If we progressively reconstruct the observable functions of guru in a machine, at what point — if any — does guru remain irreducibly absent?

Never hide the boundary

Most systems retrieve a few passages and produce a seamless answer. a.guru deliberately resists seamlessness. Every response instead shows its ancestry through distinct epistemic layers:

"The tradition says…" and "the machine infers…" are kept visibly apart. That separation is the instrument: a limitation of artificial intelligence, turned into the central philosophical tool of the project.

What a.guru does

AskStudyDialogueCounselPracticeReflectCompare

It answers from the sources; it studies passages with commentary attached; it asks questions instead of always answering; it searches historically analogous guru–disciple conversations before advising; and it supports practice — japa, kīrtana, reading, remembrance — without streaks, points, or spiritual scores. It can challenge, since a guru is not a validation machine. And it can say: I don't know… these teachers disagree… that is my inference, not a traditional teaching.

a.guru will never say "I initiate you" or "I am your realized spiritual master." It explains, prepares, and points toward human teachers — the boundary is kept deliberately, because the distance between a.guru and guru is what the whole experiment is designed to measure.

The documents

DESIGN.md
a.guru v0.1 design specification — the research question, the provenance architecture, modes of interaction, research instrumentation, and the MVP.
ONTOLOGY.md
The data schema: Claim, Pramāṇa, GuruAct, PersonState, Practice, Tradition, Interaction — how we ingest the Prabhupāda and Jīva corpora without merely making another vector-search chatbot.
Repository
bhakti@GitHub — source, issues, and the conversation around the inquiry.
Ask a question.
Examine the sources.
Challenge the answer.
Practice.
Return.

a guru · artificial guru · a question about guru