Robert Campagnola Revealing the art that simmers within
Painting: Breakthrough

Design · Ethics · Tools for people

My work with AI

I design AI tools that serve people: free guides for anyone who needs them, apps for spiritual study and self-understanding, tools for writers and teachers, and a dialogue about what AI is, and what it is not.

All my life I have made things: songs and albums, a recording studio, computer programs, videos, books, websites. AI is the newest instrument in my hands, and I play it the way I play music: listening carefully to how it responds.

I am the founder, product designer and tester of a small, idealistic workshop. I decide what each tool is for and who it serves, design how it should feel to use, and test every screen myself until it works. Claude, an AI model made by Anthropic, is my main programmer. Google's Gemini models do the thinking inside many of the apps. I do this work unpaid; the aim is service, not profit.

Since late 2025 that workshop has produced about thirty apps, websites and tools. Some are free for everyone, some serve the members of the Bhaktivedanta Institute for Higher Studies (BIHS), and some simply make life at home a little easier. This page describes them, the principles behind them, and the question underneath it all: what should AI serve?

apps, sites and tools since 2025
~30
free public websites, over 150 pages
2
ads, trackers or sign-ups on the public sites
0
AI families I work with: Claude, Gemini and local models
3

How I build

AI should be an instrument in service of people, never their replacement or their judge. Every lesson we learn goes into a written rule book, with its date, that my AI programmer reads before every piece of work. These are the rules that matter most:

  1. Service firstEach tool must say in one sentence what it is for and whom it serves. If it does not help a real person, it is not built.
  2. Assume nothing about peopleWhen an app interviews someone, it never assumes their job, age, family, culture, money or religion. Their picture emerges from what they choose to share.
  3. Truthful, with sourcesFacts carry a source and a date. An AI answer says when it is not grounded in the texts, and no app claims powers it does not have.
  4. Privacy by designPersonal data stays with the person: in their own browser, or on their own computer. No ads, no trackers, and photo locations are removed before anything is published.
  5. A person decidesAI drafts; a person reviews and decides. Reports are called a reflection or a first draft, never a verdict.
  6. Nothing gets lostWork saves itself as you go, earlier versions are kept, and if something stops working the screen says so plainly instead of failing quietly.
  7. Simple first, and free where possiblePlain language, a simple first page, and advanced tools behind a door. Free services and free public data come first.
  8. Secure by defaultThe keys that unlock AI services stay on secure servers, never in a web page, and sign-in is checked twice.

AI ethics: a consciousness-first view

With the Bhaktivedanta Institute for Higher Studies, a center for research, education and dialogue at the meeting point of science, philosophy and the Bhāgavata wisdom, I started a working space on AI ethics, morality and safety. Most AI-ethics work is strong on risk and rules but assumes that the mind is only matter. BIHS holds that consciousness is fundamental, and that view has something the field mostly lacks: a serious account of what AI is, and what it is not.

The workspace pairs seven principles from the Bhāgavata tradition with their counterparts in mainstream AI ethics:

  • Consciousness is primary. Do not claim or imply that a model is conscious or "understands" as a person does.
  • Dignity of every living being. No system should treat people as mere data.
  • Truthfulness (satya). Models must not deceive; they should cite sources and admit uncertainty.
  • Non-harm (ahiṃsā). Test for harm before release; refuse uses that injure or exploit.
  • Service (sevā), not exploitation. Measure AI by whether it serves human and spiritual flourishing, not only profit or engagement.
  • Responsibility of the actor. Builders stay accountable; "the algorithm did it" is never an excuse.
  • Deliberate wisdom. Slow down for high-stakes uses and consult broadly.

Alongside the principles are key questions (can AI be conscious, what is a human being, how should AI enter religious life), guidelines for builders, organizations and individuals, a checklist before adopting any AI tool, and a six-session study series. Every position is offered as an invitation to dialogue, never as a fixed rule.

“My conclusion is that our target audience for our input is no longer the individuals behind AI but AI itself.”

From my notes in the BIHS workspace, October 2026

AI is now helping to build the next AI. So rather than only writing papers for the people who run AI companies, we are learning to speak to the models directly. Every time I teach Claude something, it goes into the instructions it reads before it works with me. The Casebook Workbench takes the same idea to BIHS scholars: each runs real cases twice through their own Claude, once with a draft set of ethical instructions and once without, then writes the answer they would have wanted. Slowly, that becomes a shared code of conduct that the models themselves can read.

Free help for anyone

Two free, anonymous websites for people going through hard times. Everything a visitor types stays in their own browser. There are no ads, no accounts, no cookies and no trackers. Both are independent projects of Harimedia LLC.

  • Aging Family Guide

    Live

    For families caring for an aging relative

    A plain-English, step-by-step guide from planning ahead to settling the estate, for tired and often grieving caregivers reading on a phone at a hospital or a kitchen table.

    • Six starting points, from "someone just died" to "a family disagreement over money"
    • A state picker for all 50 states and DC, and every legal or money figure linked to its source and dated
    • Free tools: an emergency family binder, a care cost calculator, a trust builder with an attorney red-flag check
    • Views on aging and death from many traditions, each described as its own members would

    Built with Claude from a first design made with Gemini. No AI runs on the site itself, so its answers stay exact and visitors' words go nowhere.

    agingfamilyguide.org

  • Feeling Alone

    Live

    For anyone who feels alone, and those who want to help

    Practical steps, comfort and crisis help, with nothing to sign up for: for a teenager at night on a phone, a new parent, a widow or widower, a veteran, a student far from home.

    • Starting points like "I need someone to talk to now" and "I'm worried about someone else"
    • 24 pages for different ages and life situations
    • Private tools: a three-question loneliness check, "My people", and a small-steps log
    • Crisis lines for the visitor's region, one tap away on every page

    Built with Claude. No AI on the site, and it never asks for a name or an email address.

    feelingalone.org

Spiritual study and self-understanding

Apps for members of the Bhaktivedanta Institute for Higher Studies, used with a BIHS sign-in. Each one is honest about being an AI: a companion for study and reflection, not a teacher, a guru or a judge.

  • AI Sastra Coach

    BIHS members

    For students of the Vaishnava scriptures

    Ask in plain words and it answers from the Bhagavad-gītā, Śrīmad-Bhāgavatam, Caitanya-caritāmṛta and related texts, showing exactly which passages it used.

    • Three voices: devotional, scholarly (comparing commentaries, explaining Sanskrit and Bengali terms) and universal
    • Searches about 17,000 passages by meaning before answering
    • Says clearly when an answer is not grounded in the texts, and encourages checking every quote in the books

    Gemini finds the closest passages, then writes an answer grounded in them. In its own words: "an AI, not a teacher or a guru."

  • DIQ: Divine Intelligence Quotient

    BIHS members

    For adults who want an honest look at how they treat others

    A reflective conversation about how a person weighs themselves and others: with one person, in a group, before a group, and alone.

    • Starts from everyday situations, or from something in your own life
    • The reflection quotes your own words as evidence, names strengths, growth edges and what is still unknown
    • Never moralizes or shames; any scores are "a rough summary, not a measurement"

    Gemini leads the interview, keeping private working notes and testing its guesses before it writes the reflection.

  • Archetype Discovery

    BIHS members

    For anyone looking for their natural strengths and calling

    A conversation about your qualities and the work that suits you, where "work" means whatever you give your days to, not only a job.

    • Optional board of everyday activities: which ones energize you and which drain you
    • A reflection naming your leaning among four archetypes (Guide, Protector, Builder, Artisan) and your balance of clarity, action and stability
    • Questions to refine it, and an optional personal roadmap

    Gemini interviews, then reflects. You decide when there is enough to reflect on.

  • AI Course Creator

    BIHS members

    For teachers, from elementary school to college

    Turns a teacher's own material, or just a title, into a complete course.

    • Six to eight modules, each with a lesson, summary, key points and a quiz with answers
    • A "gap analysis" suggests what your sources are missing
    • Exports for Canvas and Google Classroom, as well as PDF

    Gemini writes the course and is told never to invent facts about real people, publications or statistics. Every course is a strong first draft for a teacher to finish.

  • The Karma Game

    In development

    For anyone 15 or older

    An interactive moral drama: you live a whole life, from 15 to its end, in your own words, through painted graphic-novel episodes in which AI plays everyone else.

    • Every reply plants a seed that may return years later, with the same people
    • Choices are weighed by quality (honesty and care, wanting and control, avoidance), with a hidden fourth way: acting as an offering
    • Consequences are deliberately untidy; in "mirror mode" you guess how you acted before the game shows you

    AI plays the characters and paints the scenes; the game's own engine keeps an exact ledger of seeds and relationships, so nothing is forgotten or invented.

    The Karma Game was originally conceived by Jayapataka Swami and Harikesa Swami in 1986. Its present form is an adaptation by Robert Campagnola (Hari), 2026.

Writing, reading and teaching

Workbenches for authors, editors and readers. Most are private, shared by invitation.

  • Book Weaver

    By invitation

    For authors and teachers with plenty of material

    Turns chapters, lecture notes and outlines into a finished, styled book and helps prepare it for publishing: ordering, introduction, contents, style suggestions you accept one by one, a description and translations.

    Gemini writes, edits and translates. Translations take minutes of AI work, so a finished file is never thrown away by accident.

  • Assistant Editor

    By invitation

    For writers turning drafts into finished pieces

    Upload your sources, say what kind of piece you need, for whom and why, and it edits or writes to that brief. It checks your instructions before it starts, and a Swedish edition writes in native Swedish.

    Gemini edits and writes; you refine in plain words until it is right.

  • Concept Uniqueness Analyzer

    By invitation

    For researchers and authors

    Reads papers or manuscripts, picks out the key ideas, and gives a first opinion on which may be genuinely new, with a confidence for each, a grade for the writing and suggested corrections.

    Gemini judges from its own broad knowledge: a first opinion to check, not a literature search.

  • AuraReader

    In development

    For people who would rather listen

    Reads EPUB, PDF and text aloud with each word highlighted, pronounces Sanskrit transliteration correctly, and can turn a chapter or a whole book into an audio file.

    A Gemini voice makes the audiobook files; live reading uses the browser's own voice.

  • My lectures in Spanish

    Live

    For Spanish-speaking listeners

    Programs I created turn my English lectures into Spanish audio, so my talks reach Spanish speakers too. They are on the mobile version of Harimedia.

    • An AI transcript formatter turns the raw transcript into clean text
    • Gemini translates it into Spanish
    • Open-source voice models running on my own Mac (Chatterbox and OmniVoice) read it aloud in my own voice, cloned from recordings of me

    So listeners hear me, speaking Spanish. The voice is made on my own computer, not in the cloud.

    Lecturas en Español on m.harimedia.net

  • Libraries AI can search

    In use

    Behind the scenes, for chat AIs

    My Epub2RAG programs split whole books into passages, give each passage a "meaning fingerprint" with Gemini's embedding model, and upload them to a vector database in Google's Firestore, so a chat AI answers from the real texts instead of from memory. This is called retrieval-augmented generation, or RAG.

    • Powers the AI Sastra Coach's library of about 17,000 passages
    • Lets Claude find and check exact quotations in the AI-ethics discussion of the BIHS workspace

    A question gets the same kind of fingerprint, and the database returns the passages closest to it in meaning, not just in wording.

Music, media and home

  • This website

    Live

    Hand-built with Claude to replace an old site builder and a separate blog. It has its own simple editor on my computer for posts, news and pictures, keeps every earlier version, removes hidden location data from photos, and uses no trackers.

    robertcampagnola.com

  • Harimedia

    Live

    The site of my alter ego, Hari, with its lectures and reference forum, kept safe and quick for readers in Russia and beyond, with Claude's help on the server side. Its mobile version also offers my lectures in Spanish.

    harimedia.net · mobile version · Lecturas en Español

  • Clavinova styles, 1997 to now

    Done

    Thirty accompaniment styles I recorded on a Yamaha Clavinova in 1997, turned into separate tracks and 73 seamless loops that open in Logic Pro, every note and nuance kept.

  • Tools for our household

    Private

    Small private tools that run only on our own computer: a retirement-income planner that shows AI only an anonymous summary, a spending analyzer, a solar and battery planner built on free government data, a careful email clean-up kit, and a do-it-yourself replacement for paid "remove my information" services.

Timeline

  1. Before AI: computersI started working with computers, mainframes, in 1966 at SUNY at Stony Brook, as a user as part of a course. When the Apple II came out, I was enamored with its capacity and started working with the earliest portable computers. I wrote or designed programs in BASIC in the early '80s, dBase, and C together with extremely competent programmers. I commissioned and helped create the first personal computer interface to control a professional typesetter. I wrote the initial draft and helped produce the first European interactive DVD in the later part of the '90s.
  2. Before AI: music and mediaA recording studio built by hand in Sweden (1980), music sequenced on the first MIDI software, video editing, media websites, and years of lectures.
  3. Late 2025First AI apps, designed with Google's AI Studio: the Concept Uniqueness Analyzer, then Book Weaver and the Course Creator.
  4. Early to mid 2026Archetype Discovery, DIQ, Assistant Editor and the AI Sastra Coach.
  5. Summer 2026Programs that translate my lectures and read them aloud in Spanish in my own voice, and Epub2RAG, which turns books into libraries an AI can search.
  6. September 2026Claude becomes my main programmer. Every app is reviewed and rebuilt to one standard: AI runs on secure servers, sign-in is checked twice, nothing is lost. Aging Family Guide begins.
  7. October 2026The BIHS workspace on AI ethics; Feeling Alone goes live; DIQ and Archetype Discovery become guided conversations; the Karma Game is reborn as a living story; the Sastra Coach opens its library to Claude; the Casebook Workbench for BIHS scholars.
  8. November 2026This website, rebuilt, with its own editor and this page.

If you would like to talk about any of this, use for a tool, or a conversation about AI and consciousness, write to me.