Morphologically, it is remarkable for having an extremely small number— perhaps less than one hundred— of basic (monomorphemic) nouns, as nearly all nouns in the language are derived from verbs.
I have long been interested in Hupa as a result of this statement. The first question we might ask is whether such a dramatic innovation is restricted to Hupa or in fact appears in some form in other lower or Pacific Athabaskan languages. There is no report of such structures in Upper Umpqua or the Rogue River languages.
Once you read the grammar and vocabulary of Hupa published by P.E. Goddard, 1905, barely half a century after their first contact with Europeans, the answer was clear. The claim about Hupa's use of verbs and paucity of nouns, which is not referenced, is totally inconsistent with Goddard's work. Goddard lists 130 nouns straight away in the first 20 pages. Not all are morphophonemic, but the non-compound morphemes for the obligately affixed nouns seem to all be unique. Furthermore there is discussion of verb nominalization on page 21-23, and while the morphology seems to be more elaborated than in Western Indo-European languages, it's nothing as dramatically novel as this statement, certainly not showing that "nearly all nouns in the language are derived from verbs." In fact Athabaskan languages in general have elaborate verb morphology, though again, they don't replace nouns.
For a time I had thought that Hupa was a real-life example of the fanciful verb-based language of Tlön fantasized by Borges in Tlön, Uqbar, Orbis Tertius. But it's not. It's worth pointing out that there does seem that there are some innovations of Hupa, relative to other Athabaskan languages, but this is to be expected from a language isolated for centuries with close trading relationships with an Algonquian language (Yurok) and a probable isolate (Karuk). Certainly the differences are not so profound.
I subsequently revised the Wikipedia article. In the meantime continue to look for languages that falsify the neurolinguistic sketch. One interesting possibility is that we might find dramatic differences in language based on geography with phylogenetic patterning. That is, is there something different about Andamanese + New Guinean + Australian languages (earliest out of Africa) vs. sub-Saharan African languages vs all other languages? I'm not asking whether there could be differences based on language-descent, which there necessarily will be; I'm asking the far more controversial question of whether they may be genetic innovations that result in different wiring and therefore differences in language structure. So far we have not found differences so profound as to warrant speculating about population-wide differences in the underlying hardware. If we ever do find differences in a language or group of languages as dramatic as the one that had been suggested here, or that Daniel Everett suggested with Piraha (which also appears to collapse under scrutiny), I submit that it might be profitable to look for anatomic and genetic differences. Such diversity of language and neurology would absolutely be a windfall to understanding the physical basis of language and cognition.
As an aside, I have been to the Hoopa Nation in Northern California several times. It's absolutely beautiful country and I highly recommend a visit. Unfortunately the Hupa language is not a living language, although it is being preserved by the efforts of Danny Ammon and others who make their resources available for the rest of us.
Trinity River south of Hoopa, California, by Trinityalpsphoto