The experience (learning) curve

node

The empirical regularity that unit cost falls by a roughly constant percentage for every doubling of cumulative production — not for every year that passes, and not for every research dollar spent. Formulated by Theodore Wright in 1936 from airframe assembly data and generalised since across hundreds of manufactured products, it is a relationship between cost and cumulative *output*, which is why it describes things that are built repeatedly on a production line far better than things that are built once on a site. The fitted rate is the learning rate: the percentage cost reduction per doubling. Photovoltaic modules have sustained a learning rate commonly fitted around twenty per cent per doubling across decades; lithium-ion cells a comparable figure; wind turbines a lower one; and on-site constructed generation frequently none at all, or a negative one. Two consequences matter for forecasting. First, a technology on a learning curve gets cheap as a function of how much of it is deployed, so deployment policy is cost policy. Second, forecasts that extrapolate cost from the present rather than from cumulative volume will systematically underestimate a technology that is scaling — which is the documented failure mode of a generation of official energy projections. Not a place.

A node is not a place. Drawing it on a map would assert something about the world that no stored fact supports.