Improvements to the algorithms for “AI Select” and “Subject Mask,” as well as much-needed options for combining masks

Sorry if I’m repeating a topic that’s already been mentioned in other posts. Some of those posts are several years old, and given how the competition has evolved, I’d like to bring this topic back into the spotlight.

The masking results from “Subject Mask” and “AI Select” can vary significantly in terms of accuracy. “AI Select” is often much more precise, while “Subject Mask” frequently produces much rougher, less accurate results. This may be due to the technical characteristics of the two masking algorithms, but the end result is still unsatisfactory.

Why?
When applying various filters, gradients, etc., I can always reference f.e. the “Subject Mask” using “Combine Masks.” That’s not possible to the same extent with “AI Select.” Yes, I can combine a mask with “AI Select” at the exact moment I apply “Combine Masks”—which I have to do right then and there—but I can’t do it afterward. When combining masks, I can’t reference existing selections, such as “AI Select” or a brushed area.
It would be great if you could combine masks not only with “Subject Mask” or “Background,” but also with other layers by their names.

Furthermore, there is an urgent need to refine the masking tools, whether f.e. “Subject Mask” or “AI Select.” Even large feathers on a bird’s wing or a leopard’s highly visible whiskers are not recognized and selected cleanly (feathers) or at all (whiskers). The spaces between the feathers are also masked. You can correct this manually, but when combined with “Subject Mask,” the manually corrected area is not taken into account.
“Refine Mask” can help, but it’s not guaranteed. It’s always a gamble whether the refinement—which applies quite broadly across the entire mask—will take the critical areas into account or not.
Lightroom set the benchmark with its latest version; it would be nice if Capture One would follow suit here.