Nuclear power has a lot of room for improvement. The problems are related to cost and complexity of the specific designs, not fundamental to the physics of using nuclear reactions for energy. Much like early computers, the current designs are large and difficult to build. There is a lot of potential to scale down and reduce costs, like was done for computers.

Molten salt reactors, for instance, offer a massive potential reduction in size and complexity (they eliminate a lot of the risks of water-cooled designs, so shouldn't need the same kind of massive containment structure to contain e.g. large volumes of highly radioactive steam in the event of a failure).

If next-gen fission research got even 10% of the resources that are currently spent on fusion, I think we would see a lot of progress towards improving the economics.

Not to disagree, but this lies in the "if we spent 5+ years we might improve in 10+ years and deploy in 15+ years" space.

If you look at the payback times on Battery, smarter networks, pumped hydro, windmill improvements, solar improvements, even now they are at their margins for 80/20 its probably shorter path to more beneficient outcome, but at a lower energy density.

The improvement in battery storage, and solar cell efficiency/cost is a good example. Over the same 5/10/15 year lifetime the drop has been continuous and at times above linear. We're now beyond the 2x improvement space, but the value of a 0.05% improvement in manufacturing for the volumes being made now, is really significant.

Nuclear, it would be very hard to project better than linear improvement in LCOE

I stress, I think we should do it. Its like the Manhatten project: Leslie Groves was asked to pick between thermal diffusion and gaseous centrifuge, and said "do both" -He was right: it turned out doing both improved feedstock quality going into the calutrons AND speed it up overall. Sometimes, its not pick A or B, its pick doing A and B and C