This is a good point. The article puts up a straw man ("one language to rule them all") when it would be just as effective to lay out the actual situation in s

This is a good point. The article puts up a straw man ("one language to rule them all") when it would be just as effective to lay out the actual situation in scientific computing -- it's uncomfortably heterogeneous at the moment. Having one language that spans more of the space from the bare metal (C-like fine-grained numerical manipulation) to the top level (say, training a complex model on a large dataset in a REPL) would be very helpful.