[johncarlosbaez]

This claimed proof is a bunch of baloney:

* First, it's written in the typical style of AI slop.

* Second, a mathematician I know and trust writes "I went straight to the technical part (Sect. 3) and randomly checked one of the results (Theorem 3.14), finding that it is obviously false. (The category Comp mentioned in the theorem is formally introduced and makes sense per se, but it is certainly not additive with the proposed definition, as claimed in the statement of the theorem)."

* Third, another mathematician I know and trust writes "I spent almost an hour poking through here carefully to see where the more central claims begin to fall apart. Theorems 3.24 and 4.1 brazenly contradict each other, proving respectively that problems in P are homologically trivial and that all NP-complete problems are homologically isomorphic to all problems in NP. Even more to the point, the proof of 3.24 really shows the lie where it says "The detailed argument uses the functoriality of the computational homology construction and the fact that homology isomorphisms preserve the 'computational topology' of problems." The last claim is, naturally, not mathematically defined. The computational chain complex also appears not to be genuinely defined, as far as I can tell. I haven't compared to see what the author chucked into the formalized definition."