But that is neither here nor there. What matters is that a provably false statement (in this case a contradiction, but that doesn't really matter) follows from some assumption, and therefore the assumption must be false. This is not circular reasoning, it is proof by reductio ad absurdum (https://en.wikipedia.org/wiki/Reductio_ad_absurdum).
There's indeed an extra assumption in here, namely that we can write a program such as B that runs on "itself". That's not difficult to see though, if we think in terms of Turing Machines (or an actual computer with memory, for that matter).
I think the given proof is essentially correct.
If you could indeed derive 1=0 from H, then that would indeed by reason to dismiss H, but you can't just "consider" 1=0, you have to prove it from H.
Simplified proof: assume H exists, then consider equation 1=0, it's a contradiction, therefore the assumption is false and H doesn't exist.