Verbatim, Most Recent, and clustered posts are all better for a mindset when you have a relatively specific query, or at least filter bias, in mind. Consider the following statements:
1) I'm looking for something specific.
2) I'd rather not be surprised right now.
3) I'd like to get through the search/index as fast as possible to get where I'm going.
The above are user-voiced versions of principles that drive traditional UI, and are absolutely better met by that optimization process.
AI-driven search and feeds were exciting because they were the first method that captured use cases where users wouldn't make those statements: fuzzy queries about general interests, the desire for diversions "curated" by your own social group, and search/feed review as itself as an entertaining experience.
Then, it turned out that these unexplored use cases were actually the giant, underwater part of the iceburg of what people wanted from information systems. The growth and attention capture you can drive from meeting those cases is not only massive, it sucks the oxygen out of traditional value propositions by out-competing for the same attention.
I think a return to traditional "capture and facilitate intent" UI (or even better a synthesis of the two!) will only be driven by users finding a way to explicitly value their own clarity of attention. We see signs of this in everything from ad blockers to focus-oriented browser plug-ins, but I think we're still missing the FOMO-conquering expression of this value proposition that becomes a competitive product.