205 billion squares of India, and no list of names

transcript

Cut India into 4-metre squares. You get about 205 billion. Now give every one its own code. Without a giant list.

Priya delivers in Dharavi, Mumbai. Her app says Building 42, Lane 7. Unmarked lanes. No building numbers. 30 minutes asking locals. Now give that spot a ten-character code. She drives straight to it.

This design uses GeoHash. Start with Mumbai's latitude and longitude. Is it in the east half of the world? Yes: bit 1. Halve again. And again. Same for latitude. Then shuffle the two together, bit by bit. Every 5 bits becomes one character. Mumbai starts T E 7 U.

Each extra character zooms in. Three characters: about 156 kilometres. Six characters: about 1 point 2 kilometres. A neighbourhood. Nine: about 4 point 8 metres. Add one check character, and that's ten.

But here's the trap. Six characters looks precise. Why not store that? Because about 30,000 tiny cells would share one code. It stops being an address. So the design never truncates.

Here's the magic. Same coordinates, same code. Every time. So a phone can work out a code offline. Pure maths. No server call. Diwali rush coming? Pre-compute every code for each delivery zone.

India Post's real DIGIPIN launched in May 2025. It splits a box over India into a 4 by 4 grid. Ten times. It needs no lookup database at all. The code is computed, not assigned.

Full breakdown in the lecture.