@Razetime I was reporting a few GNU APL bugs and noticed this discrepancy. I am suggesting internally at Dyalog that we return 0 for 0⍟0j1 but I'm not 100% certain about 0*0j1.
Well I'm not entirely sure if it would or wouldn't either way, but if you took the matrix mod 2^32 at each stage of multiplication, would it not result in something different to what it should be?
Oh oops I read the comment 'Whitespaces do not count. Only letters do. that means someone could implement it in Whitespace at a grand total of "0 bytes", right?' and assumed they were right