Integer multiplication work reportedly hits a bottleneck at κ ≈ 7.004e-4

A post about integer multiplication says the current best algorithm has a bit component and a complex component for Fourier-transform-style work; whichever has the lower “saving” determines κ. The author says recent bit-side changes have brought its saving within 0.8% of the complex side, which has held at 7.009e-4 since PR193. Further bit-side gains alone won't raise κ, the post argues; proposed solutions using conditions have not cleared the bottleneck.