I just wanted to understand my own country.

For a country this vast, our statistics leave a lot to be desired. A lot of what should be counted isn't, and a lot of what is counted somehow still doesn't count. That is the strange irony of the machinery we use to measure India. It's fine, in a way — we are still a young country — but "young" was never the same as "good enough."

Most people don't stop to appreciate this, but it blows my mind every time: India isn't really one country. It is many countries wearing a single name. Some of our states are bigger than most nations on earth; a few are bigger than several countries put together. Uttar Pradesh, all on its own, would rank among the most populous places on the planet.

And yet, even in 2026, we still don't have a reliable estimate of something as basic as unemployment. I understand why — the vast majority of the Indian economy is informal, and informal is genuinely hard to count. But hard has never been a good excuse for not trying harder.

Most of the numbers that should be common knowledge are instead trapped in dreadful PDFs, buried behind government websites that look like they were designed in 1500 BC and haven't been touched since. I'm fairly sure a child's crayon drawings have better user interfaces than some of these portals.

I've spent so many hours fighting these sites — for work, and out of plain curiosity — that I've quietly developed a theory. I'm not a religious man, but I have come to believe that being made to hunt for old PDFs behind broken CAPTCHAs and hostile interfaces is penance for sins I committed in a previous life.

This has bothered me for years, and for years I could do nothing about it, because I am not a technical person. Then large language models arrived, and I am delighted to inform the world that I am now an expert developer, programmer, coder, and several other things I don't know the correct word for. I can ship in idli-vada and JavaScript with equal confidence, mostly by talking to my computer.

Joking aside: tools like Claude Code and Codex finally let me build the thing I'd always wanted to build and never could.

The nudge came from another side project, Akshara, where I've been digitizing literary and historical works in the public domain. It gave me a front-row view of what these models — especially the vision-capable ones — can actually do. They pulled clean, structured data, tables and all, out of books from the 1800s and 1900s. It genuinely blew my mind. Along the way I met others using the same tools to rescue text from PDFs lurking behind websites that are, frankly, crimes against humanity. I figured the time I'd already sunk into this could help that cause too.

So I started with the broadest, easiest task first: gathering the high-quality numbers that are already open — Our World in Data, the UN, UNCTAD, the WHO, the World Bank, and the better government archives — before going spelunking into obscure state and municipal sites later.

Showing a chart is the easy part. The hard part is that for most normal people, reading a chart is about as intuitive as Greek and Latin. The scholars and academics already know all of this; they don't need me. So instead of just publishing charts, the whole point of This Indian Life is to write genuinely good explainers — to help an ordinary Indian, someone like me, make sense of health, money, the economy, politics, culture, and everything else that goes into living in this delightfully messy, occasionally nightmarish, and genuinely wonderful country.

Full disclosure: I lean on large language models for almost everything here, because I am exactly one person. If I sat down to collate the data and write the explainers for economics and finance alone, it would take me a couple of decades. It simply isn't possible any other way.

But leaning on these models is not the same as not knowing what's on the page. Everything here is verified and fact-checked to the best of my ability, and I'm constantly revising, tweaking, and trying to make it better. You'll notice that most of what exists so far is finance and economics — that's the corner I understand best, and I'd rather start where I can vouch for the work.

I know that right now "AI" and "LLM" are practically slurs, and anything touched by them gets treated as tainted goods. I think that's a travesty. Yes, the veins of the internet are steadily clogging with low-effort, low-intent generated slop. But — at the risk of sounding full of myself — that is not what this is.

There is a world of difference between mindless slop and actually using the language, vision, and analytical abilities of these models to do something useful. It has become a truism in certain circles that LLMs are just next-token predictors regurgitating the median mediocrity of humanity. I think that's an impoverished view of what they can do. The fact that they can help one person find good data, open it up to the public, and explain it well is, to me, proof of how genuinely remarkable they are. Without them, this project would not exist in any meaningful sense.