The article isn't saying that AI coding is "shovelware", but rather that anything which increases coding efficiency will allow the creation of more shovelware. Since there's no evidence of an increase in the rate of new shovelware being created, that's evidence that AI coding isn't a significant increase in coding efficiency.