Someone chose "most recent" when implementing "most recent", and then that decision automatically controls what is shown next. They could have chosen "most recently liked" or "most liked" or "minimum MD5 checksum" or "alphabetical" or "most liked in past 24h" or "likes / hours * followers / followee_ratio * poster_birthdate_days_diff(admin_birthday) * tfidf_rank_matching_books_published_in_1885" - all of those would be algorithmic sorts in my book. "Most recent" is just one (among multiple) of the simplest ones to predict the behavior of.

A fully human-curated feed would not be "algorithmic", in contrast. Print newspapers do this, and sometimes online ones to varying degrees, and most blogs. You could even let people build their own (e.g. "boost to my popular feed [at position X/newest/oldest]"). Essentially none do this though, they all choose algorithmic ones of varying complexity - it doesn't make them non-algorithmic to choose a simpler one.