The tools, the questions, and exactly where the AI is

People will ask — some already have — whether this whole thing is just AI. Whether a machine wrote it and I'm repeating it. It's a fair question, and instead of dodging it I want to answer it completely: here is where I use AI, here is where I don't, here is how you can tell the difference, and here is why I think using it is not a shortcut but the honest tool of my time. I'm not going to pretend the machine wasn't in the room. It was. Let me show you exactly what it did and what it didn't.

Go back to any of the old texts. Behind every one of them is a person who stood in front of something they couldn't explain and asked. The prince who saw age, sickness, and death and couldn't un-see them. The one who sat until the mind went quiet and then tried to write down what happened. The rishi with a question about the fire, the breath, the self.

And then they used the best tools they had. Memory. The guru sitting in front of them. Decades of sitting. Debate with whoever was sharp enough to argue back. Oral transmission, corrected mouth to ear across generations. The tools were the best of their age, and nobody thinks less of the Upanishads because someone used their memory instead of a hard drive.

I have a question too. Mine was small and ordinary: why am I unhappy. It grew into bigger ones — why so many gods, what is belief, why does this old sound-system feel like engineering to me. Same seat. A person with a real question, reaching for the best tools available to answer it.

The tools changed. The seat didn't.

I should tell you who's writing this, because it matters for how you read the method.

I build software for a living. I lead an engineering arm — we ship real applications, at pace, using exactly these tools: Claude, Cursor, GPT. This isn't someone who discovered ChatGPT last month and got excited. This is how I work every day, professionally, at scale. I know what these tools are good at, I know where they lie, and I know how to hold them so they help instead of mislead. That's my actual skill — not Sanskrit, not theology. Systems, and the tools that build them.

So when I turned to my own question, I reached for what I know. Of course I did. It would be strange not to. [inference]My reasoning from the sources — mine, not the tradition's.

Here is the honest description, with nothing hidden.

The AI is my library, at my fingertips, that never closes. For most of history, a question like mine meant years — finding the texts, finding someone who'd read them, sitting with a teacher, slowly. I don't have years of open time. I have a demanding job, a family, and a few hours a week. What the AI gave me was reach: I could ask a question at 4am and get pointed at the Spanda-karikas, the Devi Mahatmya, the structure of the Mahavidyas — fast, in language I think in.

It did four things, and I'll name them plainly:

• It fetched. What would have taken me weeks in a library, it brought in minutes.
• It organized. I brought raw notes and half-thoughts; it structured them so I could see them.
• It argued back. I told it, again and again, to challenge me, to be a critic, not to agree. It pushed. That mattered more than anything else it did.
• It drafted in my voice. So I could react to a page instead of staring at a blank one.

That's the library. Fast, tireless, at my fingertips. I'd be a fool not to use it, the same way a rishi with access to it would have been a fool not to use it.

This is the part that matters, so I'll be exact.

The AI did not have the question. The question came from my life — the burnout, the shiver when I sit with Ma, the thing I can feel for a few seconds and can't explain. No machine generated that. [devotional]Personal conviction, not claimed as proof.

The AI did not set the direction. Every turn of this inquiry — what to chase, what to drop, when something felt true and when it felt too clean — that was mine.

And the AI did not get the final say on what's true. This is the important one. Over and over, the machine produced something beautiful and coherent and wrong, and I caught it. I caught it importing a philosophy that didn't fit. I caught it reversing a relationship that ran the other way. I caught it giving me an answer so elegant I didn't trust it — and I was right not to. I caught it wandering out of the domain the question lived in. Every one of those corrections was mine, and the machine only knew it had drifted because I told it.

That's the difference between a tool and a source. The AI was the tool. The discrimination — is this true, is this honest, does this actually fit — was the work, and the work was mine.

I'm not asking you to take my word for it. This is the whole reason the HOW sidebars exist.

On the inquiry chapters, you can open the reasoning — and it doesn't show a clean conclusion. It shows the walk: what I tried, where it broke, where I was wrong, where I corrected the machine and where the machine corrected me. Watch me catch the AI being wrong, in real time. That's not something I can fake into existence. It's the record of a mind actually working, with a tool, and staying in charge of it.

Start with Why So Many Gods?, The Sound of the Gods, or What You're Allowed to Sing— open any "How this section was reasoned" block.

If this were just AI output I was repeating, the sidebars would be smooth. They're not. They're full of seams, because the thinking was real and real thinking has seams. [inference]My reasoning from the sources — mine, not the tradition's.

Let me be concrete about scale, because "I used AI" is too vague.

The awareness equation — the SAE — was not a prompt and an answer. It was roughly five months of conversation. Months of asking, testing, being wrong, reframing, throwing things out, starting again. The machine held the context across all of it so I could keep building instead of re-explaining. But the equation's shape — what belief is, what the variables are, what human life has that a stone doesn't — that came out of the arguing, and the arguing was mine to steer. Five months is not a shortcut. It's just a library that stayed open at 4am when the rest of the world was closed.

Here's the honest risk, and I won't hide it: the danger of a fast library is fast wrongness that sounds right. The machine can hand me something coherent and false, and if I'm not careful I'll believe it because it's elegant.

So the fast reach is not the finish. It's the start. Everything the AI helped me reach is an inference until it's checked against the actual texts — the primary sources, in their real editions, with the verses named. That checking is slow, and it's mine to do, and no machine does it for me.

Shastra is that check, built as a system. It's the knowledge graph I'm building to drag every fast-reached idea back to the primary source and ask: does this actually hold? Where does the tradition confirm it, where does it diverge, where did I just find a pattern that isn't there? The AI accelerated the inquiry. Shastra is how I keep the acceleration honest. One is the reach; the other is the ground. [inference]My reasoning from the sources — mine, not the tradition's.

And I still have to read the originals. All of them. The library got me to the questions faster — it doesn't excuse me from the reading. It never will.

If you think I'm wrong — not vaguely, but specifically — teach me. Show me where the reasoning breaks. The Respond button is on every page for exactly this. I built the whole tagging system so you'd know precisely which claim to aim at.

If you think I've fooled myself — that I've mistaken a coherent story for a true one — guide me. That's a real risk and I know it. Point at the place I stopped checking. You'd be doing me a service.

If you think I might have something — and you can help — join me. The places I've marked [gap]I looked and could not find it. What I searched and where is named. are where I need people who have what I don't: the Sanskrit, the philology, the lineage knowledge. That's not false modesty; it's a real list of things I can't do alone.

And if you think it's all just AI — this chapter is my answer. Here is where the machine was, here is where it wasn't, and here is the button to prove me wrong. Nothing here asks you to trust me. It asks you to check me.

The question is ancient, and it's mine. The tools are new, and they're borrowed. The discrimination in between — that's the work.