My argument(s) boil down to this:
One, that natural selection could not select on the basis of future or hoped-for functionality, so that this unidirectional long march from an arm to a fully functional wing on a flying bird is not possible. All you'd ever get would be a random walk around the norm for an arm and, in fact, that pretty much parallell's the experience of the people doing the fruit fly experiments in the early 1900's. All they ever got was random walks around the norm for fruit flies. Despite every effort and years of mutating and then recombining mutations in an animal which produces new generations every few days, all they ever got was fruit flies. They never got the unidirectional march to some other kind of creature, as macroevolution demands.
Two, is that the various features required for any new kind of animal could not plausibly all evolve at the same time and that, while the second such was evolving, the first would be de-evolving. In real life, when you don't use something, you quickly lose it. Thus we observe what wings look like on penguins and ostriches, which do not use them for flight.
That reality would be totally sufficient to prevent a therapod dinosaur from ever becoming a flying bird by any combination of mutations and selection.
Could some intelligent process have re-engineered flying birds from therapods? The answer is clearly yes and some of the evidence appears to support the idea.
Why did Jesus cross the road?
Because he was nailed to a chicken!
No, wait, that's Kinky Friedman.
Be right back with the correct reference.