We need a new generation of SSGs that use custom elements, and are easier to learn, with reduced complexity and simpler installation.
Agreed on the "new generation of SSGs"; disagree on the approach that relies on custom elements.
Having written tons of XUL and XBL (the forebears to Web components), I think they're great for their use case, but documents on the Web are a different use case. Documents (like this one) should still be documents—not late-bound portholes plugged by the browser. To really explore this space, XSL (or something like it in spirit) is still a better way to achieve the effect desired (even if XSL itself is gnarly and unattractive).
Hypothesis is unable to cope with annotating this page, for example, because as far is it's concerned, there is no content on it.
Build steps aren't all bad. Most of what the industry has produced in the way of tooling for the "frontend", and the threshold of acceptability those in the industry have tacitly established by use of workflows that incorporate that tooling, on the other hand, is fails to satisfy reasonable expectations.
Food for thought: consider a website that employs a static site generator that itself lives as just another piece of content available on the site—only meant primarily for the reference needs of those authoring the site content, rather than the public at large. I.e., the website as a true dynamic knowledge repository for the associated organization, and one such piece within that repository being an active paper documenting/describing how the site is put together—to such a degree of rigor that the processes and procedures described in the paper can be performed by machine, rather than needing to be carried out by a human operator.