In hard science, the null hypothesis is often 'straw-man-like', and that's usually just fine. Because, oftentimes the straw-man haunts by being kind of true, sometimes. Reality is so strange that doing en experiment that controls well enough to (near-) conclusively exclude even a trivial hypothesis is a good experiment. Not the best experiment, but a good experiment.
You don't need a null hypothesis to do science. If you want to measure the strength of the Moon's gravitational pull, no-one's going to complain that you haven't outlined a null hypothesis.
There's all kinds of good science done that doesn't use the null hypothesis - it's a myth that if one isn't presented, then it's "not science".
You are talking past his point, perhaps willfully...
They are saying that in hard science it is often useful to start with a null hypothesis, and report anything that diverges from that as potentially significant (assuming statistical significance of the divergence).
Hell, given the null hypothesis is an assertion of perfect equality, I'd argue it's inherently a straw-man. It's one of the reasons behind "Fail to reject H0" rather than "Accept".
