Suppose Ax = b always has at least one solution no matter what b is. Why AT y = 0 has only the trivial solution y = 0?

For the first equation we have:

A*x = b

solving for x, we get:

x = b/A

If it always has a solution, then we can not have A = 0, because that causes an undefined operation.

so for example, if we have A = 1 and b = 2

x = b/A = 2/1

For the other case,

A*y = 0

dividing both sides by A

y = 0/A = 0

y = 0

Here we have only one possible solution, the trivial one, y = 0.

And the dependence on A disappears (because the quotient between zero and a number different than zero is always zero)