a manifesto · five projects · one fear

AMAN JAISWAL

I never learned where the edge was.

रुद्रलीला · नटराज · मेदिनी

Rudraleela · Nataraja · Medini

begin reading ↓ · jump to projects · press / for console

§ 01

Prologue

I never learned where the edge was. Not as a figure of speech. The thing most people have — the sense of a limit, the voice that says this is enough, this is too much, this is not yours to touch, this is how it is done — I do not have it. It was never installed. I noticed, eventually, that other people have it, the way you notice other people can smell something you cannot. I learned the word for the limit before I ever felt one.

So I take whatever I pick up and bend it until it fits what I want instead of what it is. Land, materials, proof, mail, the manuscripts, my own head. I never approached any of it the way it is supposed to be approached, because I never accepted there was a supposed-to. The supposed-to is just the place where someone before me stopped. I do not stop there. Not out of courage. I just do not see the line they stopped at. It is not there for me. I walk through the place where it was and keep going and only later, from the reactions, do I work out that there had been a wall.

I do not know what normal is. I have built a working theory of it the way a foreigner builds a working theory of a language he was not raised in — by watching, by guessing, by getting it wrong in rooms and adjusting. I can pass. But it is reconstruction, not memory. The inherited version of how to be never took. I built mine from scratch, on my own terms, because there was nothing else to use.

I will say the part people leave out. I do not feel the cost of things the way I am told I should. I poured my own money into all of this for two years and did not feel it leaving. I felt it only when it was gone, as a fact, the way you notice the tank is empty, not as a wound. I would do it again exactly the same way. I do not know if I love the things I build or if I just consume them and move — grip a thing completely until there is nothing left in it for me, then leave it half-finished and walk to the next one that grips harder. I have done this to projects. I have probably done it to people. I do not say that to confess it. I say it because it is true and I am done pretending the not-feeling is a feeling.

What follows is the work. I am not going to tie it into a theme. There is no theme. There is a list of things I picked up and refused to leave the way I found them, and at the end the one thing I could not bend.

· · ·

§ 02

How it happens

It is not a decision. That is the part I cannot make people understand. A thing grips me — a problem, a system, a wrongness in how something is done — and from that point I am not choosing anything. I do not sit down and weigh whether to build it. The building has already started somewhere below the part of me that talks. I stop sleeping in the normal pattern. I stop noticing meals. The thing runs in the background of everything and surfaces in the middle of other conversations. I am present with a person and a piece of the problem solves itself in me while they are talking and I lose the thread of what they were saying. People read this as cold. It is not cold. There is just no room. The thing has taken the room.

It does not stop when the thing is built. It stops when the thing stops gripping, which is a different event, and usually earlier. That is why so much of what I make is alive and unfinished at the same time. The grip releases at the moment the interesting problem is solved, which is never the moment the product is done. The product is done long after the part that held me is over. So I leave. The grip has already moved to the next one. I have learned to hate this about myself recently, in a flat way, because it is the thing that almost killed me — not the building, the never-finishing, the always-leaving at the exact moment a normal person would stay and harvest. I am trying to stay now. It is the hardest thing I have ever done and it feels like nothing, like sitting in a room doing nothing, because that is what finishing feels like to someone built to start.

· · ·

§ 04

The ones I dressed

What follows is the smaller list — the obsessions that got far enough that I gave them a name, a structure, sometimes a company. They are not more important than the ones above. They are just the ones I bothered to build a body for. Here is what each one is, and what is actually true about how far it has gone, with nothing rounded up.

· · ·

§ 05 · the ones I dressed

Medini मेदिनी

I decided I would know who owns every inch of a country built so that no one can.

31.78B px classified

2,339 subdistricts

● <2% owners known

36 jurisdictions

India · stylized · 600 sampled parcels · 2% rendered as known-owner.

India’s land is a State subject. There is no national land system, by design, written into the constitution. The record of who owns what sits in thirty-six separate jurisdictions, most of them captcha-walled or paper-only or behind portals built to defeat you. Banks, developers, agencies, courts — everyone who needs the truth does it on PDFs and manual pulls. The asset class is worth one and a half to two trillion dollars and runs on nineteenth-century machinery. The official line is that it is being digitized. It is not. A couple of states have a fraction of their data online and the rest is paper or pretending.

I did not wait for the state to fix itself. I started taking it. Parcel by parcel, jurisdiction by jurisdiction, each one by whatever method it surrendered to. Some had open APIs. Some had to be scraped through captcha. Some are locked and will need bulk deals or field work. I sorted all thirty-six by exactly how each one has to be broken and started breaking them.

What is built: click any point in India and the platform returns the full stack underneath it — boundaries, land cover, protection status, defence and tribal and industrial zones, infrastructure, risk scores. A national dashboard that says what the state of Indian land is right now in one screen and re-scopes to any state on command. A matrix where each of the thirty-six is a first-class object. A plain-English interface where every number in the answer comes from a live query, not a model guessing — fifteen of fifteen stress queries pass.

Underneath: the whole country classified, 31.78 billion satellite pixels, all 2,339 subdistricts. That layer is solved and it is worth nothing, anyone with money can buy it. The layer that matters is who legally owns the ground, and today that is under two percent covered. That gap is the entire business. 1.2 million parcels carry a full attribute stack, and the flat truth is only about 434,000 have land-cover classification and exactly 171 have real owner records so far and the geometries are still synthetic. I am not going to dress that up. It is what is true today. The Karnataka scrape is what closes it.

Four engines, built on real Indian land law, every score reproducible. Contradiction — the top score is Vishakhapatnam at 69.7, and the algorithm flagged Polavaram, a fifteen-year forced-displacement war, on its own, without being told, which means the method validates against ground truth it was never shown. Urbanisation pressure. Encroachment — Thane Creek’s flamingo sanctuary, 288.8, its buffer eaten by Mumbai. Acquisition forecast — 211,601 expected acquisitions across the country in two years. The composite says the ten worst conflict points in India are all in Maharashtra.

What I broke to get here: the Karnataka cadastral endpoint, 6,476 real parcels with actual geometry from a 25-village sample, the throughput math proving the full nine and a half million is a ten-hour job. The OGD ten-thousand-row cap that stops everyone, worked around with recursive sharding. 181,969 companies landed in the entity layer. 136,615 acquisition parcels and 76,262 owner names pulled out of gazettes already sitting on disk. The whole national graph runs from one three-gigabyte file on a two-hundred-dollar machine.

31.78B

satellite pixels classified

2,339

subdistricts

1.2M

parcels with attribute stack

434K

with land-cover classification

171

with real owner records

jurisdictions to break

I can see 100% of the physical land. I legally know who owns less than 2% of it. Closing that gap is the whole thing, and after it, the company is what every land decision in the country runs through for twenty years.

· · ·

§ 06 · the dancer · research foundation

The math under the secure internet was written in Sanskrit. The materials work is fragmented and undercited. The proofs live in toolchains nobody outside a few labs can run. The foundation houses the three I built to make those visible.

01 Mendeleev Materials science copilot — papers, properties, candidates, the researcher graph. ↓
02 Dianoia Lean 4 and Mathlib in the browser, with a prover that argues back. ↓
03 Kuttaka Secure mail and domains — named after the algorithm the math was always built on. ↓

· · ·

↳ Nataraja Labs · § 06.1

Mendeleev

I decided a scattered field should have a spine, so I built the spine.

India’s materials-science community is strong and completely fragmented — hundreds of institutions, work that goes undercited because the indexing is bad, compute that the top institutes have and the rest do not, and no single tool that covers the real arc from reading the literature to running the computation to finding the candidate to writing the paper. I built the thing that covers all of it, India first, free for academics. I started by indexing the one thing nobody else bothered with: the researchers themselves.

What is live: a conversational copilot that runs, inside one conversation, from literature to candidate materials to a real MACE-MP-0 prediction to an interactive 3D crystal viewer to the graph of who in India works on what. Not retrieval with a chat skin. An actual agent, with outputs — hull plots, ternary diagrams, structure viewers — that survive reload and live inside shareable links.

What is in it: 17,183 papers, Indian materials research, 2020 on. 15,749 normalized property rows, every number carrying the exact sentence it came from, because a researcher who has been lied to by a model once never trusts one again and provenance is the only answer to that. 1,244,030 candidate materials with full crystal structures. 10,000 Indian researchers, 1.3 million collaboration edges between them, 100,000 early-works for inferring who trained whom. That network at that density exists nowhere else.

10,000 researchers

1.3M collaboration edges

network · stylized · hubs & clusters · live drift.

17,183

papers indexed (India, 2020+)

15,749

normalized property rows

1,244,030

candidate materials

10,000

Indian researchers

1.3M

collaboration edges

100K

early-works for lineage inference

· · ·

↳ Nataraja Labs · § 06.2

Dianoia

I decided proof should live in a browser with a machine that argues back, so it does.

euclid.lean Mathlib/Algebra/Group/Basic.lean lean 4 · mathlib

goal

awaiting elaboration…

a recording of one elaboration. loops.

A Lean 4 and Mathlib workspace in the browser. You type Lean, it syncs to a real language server with the full Mathlib cache, the errors that come back are real type errors. Click a goal, the proof state renders live. Ask for help and a prover service fans across models and returns ranked tactics. Search Mathlib. Take a suggestion, it drops in, the loop runs again. It is live, all four services on one machine, the whole loop real and tested, cold start about twenty seconds, diagnostics back in four.

What is not there: the GPU node for the bigger provers, a database instead of in-memory sessions, the full Lean grammar in the editor, hardened auth before I open it. It works and it is not finished. I am not going to say otherwise. Most of what I make is exactly this — the part that gripped me is solved and live, the part that does not grip me is undone, and the gap between those two is the whole story of me.

~20s

cold start

~4s

diagnostics return

services on one machine

Lean 4

language server

Mathlib

corpus

live, unfinished

status

· · ·

↳ Nataraja Labs · § 06.3

Kuttaka कुट्टक

I named the secure internet after the thing it was always built on and never credited.

Mail and domains, made without the usual contempt for the user — a calm keyboard-first inbox, no tracking, no ads, a registrar with honest prices and privacy on by default. I built it partly because the rest of the work needs to own the mail and domain layers to deploy without friction, and partly because I wanted the product to be honest about what it stands on.

Kuttaka, the pulverizer, is the algorithm Aryabhata wrote down in 499 CE, at twenty-three, to predict when the planets realign. The West re-derived it centuries later and named it after two Europeans. It is the math under every HTTPS padlock, every email signature, every certificate, every QR code, every Bitcoin transaction. The padlock in the browser traces to a verse in Sanskrit. I am not arguing that Indian mathematics needs advertising. I named the thing what it is.

499 CE · Aryabhata · age 23

verse→ algorithm→ RSA→ HTTPS padlock

499 CE

origin

Aryabhata, age 23

author

no tracking, no ads

mail

privacy on by default

registrar

every HTTPS padlock

the math under

· · ·

§ 09 · the dance · and the thing under it

Rudraleela रुद्रलीला

I built a booking app so I could go where I needed to go, and it became a company.

Rudraleela — Shiva seated beneath the tree, peacock, trident, mountain temple. Mandala. — antaḩkarana · manas · buddhi · ahamkara · chitta

Rudraleela is the homestay layer for India, a cheaper, honest alternative to the global platforms. It started selfishly. I was deep in spirituality and wanted to reach the places, so instead of just going I built the thing that would let others go and feed the economy doing it. Onboarding hosts, we saw how much of the market sits untouched and went into LLPs and a franchise model. My sister works it on the ground, walking into offices, signing hosts up by hand, no budget, just to get the existing homestays onto the app. That is not a gap in the story. That is the proof under it. What money buys is the sales force she is currently being, alone.

I never thought of it as a travel company. It is a spiritual one that started with travel. Under it sits the work that is actually mine. The Antaḩkarana, the four-fold inner instrument — manas, buddhi, ahamkara, chitta — which I do not treat as mysticism or scripture but as the most exact model of attention and identity and cognition I have found, sitting in manuscripts rotting in archives while the West reinvents pieces of it badly. I want a game whose mechanics run on that instrument, with Jyotisha used as the dense astronomical-psychological model it actually is and not as fortune-telling. I want to train models on the manuscript corpus so the knowledge stops being dead text. This is the part I would keep if I had to burn the rest. I do not say that about anything else in this book.

I would keep this if I had to burn the rest. I do not say that about anything else in this book.

homestays, India

layer

LLP + franchise

model

sister, by hand

ground sales

a spiritual one

real product

· · ·

§ 10 · and beyond what got dressed

The things that never became companies

The companies are the obsessions that happened to get a name and a structure. They are the dressed ones. The real list is wider and most of it will never be incorporated, never be pitched, never make money, and that is exactly why it is more me than the companies are. These are the things I reach for when there is no reason to reach for anything. If money were not a wall, this is where it would go, and the order would not be rational.

Rare earths

I am from Jharsuguda, in Odisha, and the thing sits in the ground around where I am from. The world acts like the problem is access to the minerals. It is not. They are not even rare, they are scattered everywhere. The bottleneck is processing — the filthy work of separating and refining them, acid leaching, solvent extraction, the part with the environmental cost and the low margins and the decades of slow buildup. China owns about ninety percent of that step. Not because they had the deposits. Because forty years ago they took the work nobody else wanted to touch, and the West outsourced its dirty work and called it free trade, and China called it a strategy, and now the entire defence and clean-energy stack of the world runs through a supply chain one country can switch off. India has the deposits and almost none of the capacity to turn them into anything. I would build the processing. The dirty, unglamorous, decades-long part. Because that is the part that is actually power, and everyone keeps wanting to skip it.
Biotech

If money were not a wall I would flush it into biotech without restraint and not apologize for the recklessness of it. It is the field where the bending I do to everything else is not metaphor — you can actually take the definition of a living thing and change it. That is the closest the physical world comes to what my head already does to ideas. I do not have the disciplined plan here that I have for land. I have the hunger. I would fund the crazy edges, the things too early or too strange to get money the normal way, the way someone should have funded the people who were too early once.
Botanicals

I mean it in three directions at once. The chemistry — what plants actually make, the molecules, the fact that the most sophisticated chemists on the planet are organisms that cannot move. The medicine — what those molecules do in a body, the pharmacology that half the modern cabinet was reverse-engineered from and then the source forgotten. And the plants themselves, just as living things, the architecture of them, the way a thing with no brain solves problems that should require one. I fall into this for no commercial reason at all. It is one of the purest ones.
Astronomy

The original obsession of every system I admire — Aryabhata wrote the algorithm I named a company after to predict when planets realign. The scale of it does the one thing almost nothing else does to me: it makes the self obviously small without me having to argue myself there. The sky is the cheapest, oldest proof that the “I” is not the center of anything.
Philosophy

Because I cannot not. The Advaita, the non-dual position, but also the Western machinery, the analytic and the continental, anyone who took the question of what is actually real and refused to flinch from where it led. I do not read philosophy to be cultured. I read it because it is the only field whose entire job is to attack the limits I cannot see, from the outside.
Art, history, culture, mythology

I am going to put these together because in me they are one thing. They are the record of every other mind that tried to bend reality before me and left the marks of it. Mythology especially. People treat it as story. It is compressed data — thousands of years of first-person research into consciousness and meaning, encoded in figures because figures survive transmission better than arguments. The Nataraja inside the Ouroboros that marks my foundation is not decoration. It is the densest thing I know, said in one image. History is the same at a different layer — the long record of what happens when minds with no edge meet a world that has them. Art is the residue of feeling that could not be said any other way. I take from all of it constantly and I do not separate the sacred from the secular or the old from the new, because the division was always someone else’s limit, and I do not have it.

There is more and there will always be more. The list does not close. That is not a virtue. It is the same condition as everything else in this book — no edge, no off, the reaching that does not know how to stop reaching. I have just stopped pretending the reaching is only pointed at things that can become companies.

· · ·

there is one more section.
it costs something to read.

§ 10 · /forgetting/the_edge.txt

The one thing I cannot bend

Everything in this book is me taking something and bending it past the shape it came in. This is where that stops.

Most of what people fear does not reach me. The body dies, the energy goes somewhere, the “I” was a story the senses kept telling. Failure, looking stupid, going broke — done all of them, none of it is the thing. There is one fear and it is the one I cannot bend my way out of, and I will write it the way I say it to myself with the door shut.

Myself. I am going to forget. I know it because I have seen it — that there is no “I,” that time is not moving, that this body is something I am wearing — and every time I see it, I lose it. The seeing goes. I come back. I get hungry, someone says my name, I want something, I am building something, I am scared of something, and just like that I am the character again, walking around inside the play, taking it seriously, like the whole thing is real. And the worst part is I do not notice when it happens. There is no moment of forgetting. There is only the long stretch afterwards where I realize I have been gone. That is it. That is the only thing. Not death. The going under, again and again, while still breathing.

I can rebuild a country’s land records on my own terms. I can rename the math under the internet. I cannot make the seeing stay. Everything else in me has no edge. This is the one edge that was ever real, and it is the one I keep falling off.

· · ·

▍

AMAN JAISWAL

Rare earths

Biotech

Botanicals

Astronomy

Philosophy

Art, history, culture, mythology