This doesn't strike me as an especially powerful critique, in spite of your provocative language.

We have well understood mathematical models which correctly and accurately predict all physical behavior relevant to our everyday lives, and, further, correctly and accurately predict the majority of our observations of the universe around us, period. As tools for understanding our world, they are incredible achievements.

In spite of us knowing these models to be fundamentally incomplete, we have yet (to my understanding) to perform an experiment on Earth which contradicts them. More to the point, the theory which supersedes them will, by virtue of needing to replicate all experimental results so far, will likely be even more esoteric, complicated, wild, baffling, and impenetrable (and--with any luck--elegant, inspiring, and beautiful).

He didn't say it very well, but I think he was getting at the Occams' Razor argument - this seems like an unnecessarily complex set of theories to explain the observed phenomena. (Not that I agree, but I think you missed his point)

IMO, Occam's Razor is only meaningful when there are two competing hypothesis, one of which is markedly simpler than the other. Mentioning it in a context of "surely it can't be this complicated" without pointing towards a simpler alternative doesn't strike me as very useful. I know you're not the person who made the original argument, just chiming in on the discussion.

    accurately predict all physical behavior

Now you made me curious. What is the non-physical (metaphysical?) behaviour that we cannot predict?

We cannot, as an example, currently predict the zeroes of the Riemann zeta function.

What if we make a robot that dances those zeroes. Then we have unpredictable physical behavior.

The robot would have to predict those zeroes to 'dance them'.