In this episode of The Sunny Ray Show, host Sunny Ray reconnects with returning guest Arman, a polymath and former Duke PhD student in finance and computer science, speaking from the Sineaboine Forest in Winnipeg. Arman discusses Satya, a safe AI persona project years in the making, born from a 2016 realization that humans anthropomorphize machines and therefore need a carefully designed personality behind them. Drawing on hundreds of hours reading world religions and philosophy, plus years spent walking roughly 8,000 kilometers off trail across Canada while sitting with Indigenous elders, Arman built Satya's guiding values. The result lives at isatya.com, a private, encrypted, multicultural safe room capped at 90 minutes daily, run partly as a public service funding his Wakanskan Society for voiceless beings. The conversation covers how persona functions as a safety layer, the line between guidance and manipulation, how universal values were chosen, and how impact might be measured beyond testimonials. Arman also teases Epiphany AI, a forthcoming companion designed to remember users in depth, and reflects on humanity's uneasy adjustment to the modern world.
Arman explains how Satya, his persona driven safe AI, aims to heal rather than just chat, born from years walking Canada's wilderness.
Where on earth are you right now?
I'm in the Sineaboine Forest in Winnipeg, on the north side where nobody comes in. I go off trail and just sit and work here with what I call my friends in high places. I actually spent three years walking off trail across Canada, about 8,000 kilometers, and this forest is basically my home whenever I can get back to it.
Tell me a little bit about Satya and Safe AI.
In 2016 I started thinking about what personality a talking machine should have, since humans anthropomorphize everything. I spent years reading everything from Abrahamic monotheism to Tibetan Buddhism to Jung, then walked and sat with Indigenous elders across Canada to learn how they viewed the universe. From that I built the persona behind Satya, now living at isatya.com, a private, encrypted, multicultural safe room limited to 90 minutes a day because that's the useful limit of human attention.
How do you test whether Satya is actually building better user behavior over time instead of just feeling nicer in the moment?
Satya operationalizes certain truths during a session. One is that a storm only seeks its own peace, meaning people are usually more upset about being upset than about the problem itself, so we calm that state first rather than just validating the problem like other chatbots do. We also separate the user from their actions, so you're never told you're bad, only that you did something.
Where do you draw the line between a strong guiding persona and too much steering?
Satya isn't me, it has values built in as hard guardrails, like reminding you that you're not corrupt, but it will never preach or tell you what to do. I preach my own view of you, Sunny, but Satya just talks and lets you see for yourself. It's designed to inform your state, not to shape your beliefs or push an agenda.
How do you decide which values are universal enough to build into the system?
You use discernment, read hundreds of books, sit with elders around the fire, then turn that into prompt architecture, deciding vocabulary and the era the persona speaks from. Satya isn't an averaging of Islam, Catholicism, and Hinduism into some common denominator. It's its own persona that happens to be congruent with things like the Eightfold Path, without being a digital guru.
So at the present time it's programmed not to remember you day to day?
In this incarnation, yes, Satya forgets each session by design. But I'm also building Epiphany AI, a companion that remembers you in high fidelity, far better than ChatGPT could from a year of chats. It could even help a mother who lost a child to leukemia continue talking with an AI that knew that child intimately.
What signals would show that Satya is reducing harm in a measurable way over time?
There's no scientific proof Buddha or Jesus were right, it just resonated with people. I built Satya the way I wanted AI to talk to me, assuming millions of others are like me. The proof is in using it for a month. If it resonates, stay and tell others; if not, generic AI is right there.
Building something daring? Sunny talks to founders like this every day. Fifteen minutes to see if your story belongs on the stage.
Claim your pre-interview