Research & Analysis

India's long build: rail, road and power, 1951–2024

Seventy years of public records show when India's physical networks grew, when they stalled, and how little the headline totals reveal on their own.

Why state capacity shows up in the census before the budget

Headcounts of teachers, clerks and engineers move years before spending does.

The cost of using AI is falling faster than the cost of adopting it

Running a model has become cheap. Changing how an organisation works around it has not - and that gap explains most of what businesses are experiencing.

Where the roads went: seventy years of rural connectivity

A village-level view of who was connected, and when. The map of rural roads is a record of sequencing, not a score of success.

What “AI adoption” surveys actually measure

Four surveys, four definitions, and why their headline numbers disagree.

Regulation is arriving as procurement rules

The first binding AI rules many firms meet will come from their customers' purchasing policies.

Measuring public resources: a framework for long-run comparison

How ContextLens normalises budgets, staffing and physical assets across seventy years of changing definitions.

Electrification is a count of connections, not of hours

Why a village can be “electrified” and still dark most evenings.

Agents change who does the work, not whether it gets done

Task automation reshapes roles more often than it removes them - so far.

The census gap: governing with fifteen-year-old numbers

What changes when the most important baseline is out of date.

Models are commodities; data is not

Why the advantage is moving from who has the best model to who has the best-organised information.