It’s made serious engineering progress too.
There is no possibility of any present scheme operating at even 10x the cost of fission. Fission is not today competitive, and falls further behind each day.
We've made massive strides compacting designs, thereby transforming their unit-economic envelope, using low-temperature superconducting magnets. Those magnets continue to improve, driving potential gains in designs faster than experiments can be funded and built. Optimizing for structural materials, or even blanket versus replaceable structure, seems premature when we don't know the parameters or even type of bombardment we'd be working with.
Period.
It might be, in a hundred years, if our present understanding is wrong, and somebody figures out how to reflect gamma rays.
Their scheme has been soundly criticized in the literature as not working by a factor of ~100.
Another alternative is magnetic confinement, but radiative loss goes up with the 4th power of temperature, so would be 10000 times as much as for a D-T plasma, IIUC.
https://hardware.slashdot.org/story/12/04/11/0435231/mit-fus...
We could compare it to the state of general AI, but at least in the machine learning field, progress is being made without knowing the feasibility or path to reaching general in AI. In fusion it appears that the theory has already been set and maintaining the chain reaction going for long enough is the limitation (progress being made here), would it be the same case that the final steps are still missing without a clear path, or would the current progress be enough to eventually reach it?