The crux of the proof is on p. 3:

I have tried it with high numbers and it seems that there is no problem, which would be to say that there are infinite prime numbers separated the quantity that we want.

I'll note that the method requires taking large symbolic derivatives and then factoring integers far larger than the twin primes generated, so this is not an algorithmic improvement.