What It Is, and Where It's Going

Shastra is the part of this project that keeps the rest honest. It's where a claim gets checked against the actual texts and either earns its keep or gets thrown out. It's also the hardest, slowest part — mostly unbuilt, a lot of work ahead, and the place I most need people who know things I don't. This page is an honest account of what it is, how it works, and where it stands right now: early.

A normal reference layer is a list of citations. Shastra can't be that, and the reason is structural.

The knowledge I'm working with was never gathered into one place. It lived scattered — across sects, lineages, regions, texts written for people who already had the context, teachings passed mouth to ear and never written down. No one ever held all of it. So a flat citation list would quietly lie about the shape of it — it would pretend there was one settled body of knowledge to cite from, when there never was. [inference]My reasoning from the sources — mine, not the tradition's.

To represent something distributed honestly, the structure has to be distributed too. So Shastra is built as a knowledge graph: nodes (what each text or lineage actually holds) connected by edges (how ideas relate across lineages that never spoke to each other). And the edges — the connections no single lineage could draw, because none of them saw the whole — are the actual contribution. Not new claims. The connective structure, made visible. [inference]My reasoning from the sources — mine, not the tradition's.

This is the discipline, and it's the same one running under every chapter on this site. A claim moves through it like this:

1. It comes in — usually as something I noticed, reasoned, or was told.
2. It gets sourced — checked against the actual texts, in order of authority: primary text first (the original, with the verse named), then critical scholarship, then devotional compilations, then secondary summaries, and last, honestly, my own lineage knowledge or reasoning.
3. It gets tagged — by exactly how well it held up. [attested]A named text, edition, and verse — linked. If it isn't linked, it isn't attested. if a named text supports it. [inference]My reasoning from the sources — mine, not the tradition's. if it's my reading. [contested]Specialists disagree. I name the sides instead of picking the one that flatters me. if scholars disagree. [gap]I looked and could not find it. What I searched and where is named. if I looked and couldn't find it.
4. It's held or dropped — and this is the part most projects hide. If a claim doesn't survive the check, it doesn't get quietly kept with soft language. It gets downgraded, or removed, and the fact that it failed stays visible.

The tag has to change the claim. If I write something as [attested]A named text, edition, and verse — linked. If it isn't linked, it isn't attested. and can't produce the text, that's not a small sin — it's the one thing this whole project exists not to do. A [gap]I looked and could not find it. What I searched and where is named. honestly admitted is worth more than a citation I made up.

Here's a real one, because "open to correction" means nothing without an example.

Early on, working out where the goddess Kamakhya sits in the structure, I had a clean, elegant answer: she's the single source-point, the bindu, the dimensionless origin every other form emanates from. It was tidy. It felt right. It was the kind of answer you want to be true.

I dropped it. Not because someone disproved it — because it was too clean, and "too clean" is a warning light, not a green one. When I actually pushed on it, the truer answer was messier: she has no single seed or image not because she's one point, but because she's the source that contains all of them — the origin can't be depicted as one of its own products. The elegant version got thrown out. The messier one, which I can actually defend, stayed.

That's the method working. The elegant answer died. I'd rather show you the graveyard than a spotless lawn. [inference]My reasoning from the sources — mine, not the tradition's.

I won't dress this up. Shastra is mostly unbuilt.

What exists is the design — how the graph works, what a node needs, what an edge needs, the rule that every edge is itself a claim and carries its own confidence and its own falsification test. What exists is a source hierarchy — which texts count as primary, which as secondary, so nothing lower ever quietly stands in for something higher. And what exists is one small worked example, built by hand, to prove the method works before I scale it.

What doesn't exist yet is the hard part: actually working through the primary corpus, text by text, verse by verse, building the nodes and drawing the edges with the sources named. That's slow. It needs the languages I don't fully have. It's the difference between designing a library and actually reading every book in it. I'm at the design, standing at the door of the reading. [gap]I looked and could not find it. What I searched and where is named.

Working surfaces that already exist: source index · observation 3.8 · How This Is Made.

This is where I need people, and I'd rather say so plainly than pretend I can do it alone.

If you have the Sanskrit, the philology, the training in these texts that I don't — the places I've marked [gap]I looked and could not find it. What I searched and where is named. are a real list of things I can't finish by myself. If you know a lineage from the inside, your knowledge is a kind of source I can't get from a book. And if you just think the whole approach is wrong, that's useful too — tell me where, and how you'd check it.

I'm not asking anyone to believe anything. I'm asking for help building a thing that checks itself honestly, and telling you truthfully that it's early, it's a lot of work, and I'm one person doing it in the hours I have. The Respond button is on every page. The [gap]I looked and could not find it. What I searched and where is named. tags are the map of where I'm stuck.

The reach to information got faster in my lifetime — that's how I got this far this fast. But the reading, the sourcing, the slow grounding of it — that doesn't get faster. That's the work. And that's what Shastra is.

If you want to help, or challenge, or just watch it get built: Respond, anywhere.