Technology that moves fast

This is Part 2 of a series on building an open-source Indian address parser.

How we took 4.37M raw MCA and bank address records with zero labels to a fine-tuned, benchmarked, publicly-published LLM — and why most of the hard problems weren't in the model at all.

How Peace, Prosperity, and Evolution Are Quietly Ending Human Civilization As We Know It

A first-principles exploration of what separates biological neural computation from modern LLMs — and the decades of research that tried to close the gap.