It's just impossible to draw the kind of line between birds and dinosaurs that should be there if the split happened as far back as he puts it.
From the article:
How can a bird hand, for example, with digits two, three and four evolve from a dinosaur hand that has only digits one, two and three? That would be almost impossible.
Dueling impossibilities!
So far, the only embryo evidence for the article's hypothesis is from ostrich embryos. Suppose the next embryo experiment with another species shows just the opposite.
Perhaps there are two totally different lines of birds extant today?
Some conflicts come from bad data. In such cases, further study typically resolves the conflict, only to have a new one spring up nearby.
There's some basis for lumping birds in two piles but it seems to be a beauty-contest judging call.
So far, the only embryo evidence for the article's hypothesis is from ostrich embryos. Suppose the next embryo experiment with another species shows just the opposite.
There's a published study by Hans Larsson and GÜnter P. Wagner
Pentadactyl ground state of the avian wing
The issue of the homology of bird fingers with those of pentadactyl amniotes has been a topic of contention for nearly 200 years. Data from the fossil record and phylogenetic systematics ascribe bird digit homologies to digits I, II, and III of pentadactyl amniotes while embryological evidence supports digital homologies of II, III, and IV. Using a molecular marker specific for condensation competent mesenchymal cells, we describe a pentadactyl arrangement of prechondrogenic digital anlagen in the wings of stage 29 chick embryos. Only the middle three anlagen develop into mature fingers. This pattern supports the hypothesis that bird fingers develop from digital anlagen II, III, and IV of pentadactylous amniotes. In addition, this result rejects a model assuming a shift in the primary axis in bird digit development and shows that a prechondrogenic digital anlage has been maintained in the bird lineage for at least 220 million years since the last known pentadactylous ancestor of the lineage. Such a vestige suggests that strong constraints are maintaining a pentadactyl ground state in amniotes. J. Exp. Zool. (Mol. Dev. Evol.) 294:146-151, 2002.
I don't think that a shift in digit selection under control of a gradient is a difficulty. But the fossil sequence needs to show it, too. Not knowing dinosaur development, the digital anlage is an open question, however.