Kerala Rising: From Exporting People to Exporting Ideas · Page 96 of 353

KERALA RISING · FROM EXPORTING PEOPLE TO EXPORTING IDEAS

KERALA RISING · FROM EXPORTING PEOPLE TO EXPORTING IDEAS
of a technology programme. Most systems get more
expensive to run over time, as scope creeps and
maintenance compounds. This one gets cheaper,
because its core operating cost, inference, is on a
steep downward curve. Equivalent the system
capability has been getting roughly ten times
cheaper each year, which means a program
designed against today’s prices will run on a fraction
of that cost within a few years, even as volume
grows.
[5]
This changes how a finance reviewer should think
about the long horizon. A conventional technology
cost projected forward compounds upward and
eventually dominates a budget; this one, projected
forward, shrinks toward zero as a share of operating
cost. The people-and-coordination costs; the Keralabased team, field operators, audit, are real and will
grow modestly with the programme, but the AI line,
already a rounding error, becomes a smaller
rounding error every year. A reviewer modelling a
pessimistic case does not need to worry that the AI
cost might balloon; the structural pressure runs
entirely the other way, and the only honest risk to
model is on the revenue ramp, not the compute bill.
The same trend quietly widens the range of
behaviours the machine can afford to verify. A civic
action that is marginally too costly to verify at scale
today, because it needs a frontier model on every
submission rather than a cheap classifier. Becomes
96