Closed GinoGiotto closed 2 months ago
A shortening:
abv $p |- ( { x | ph } = _V <-> A. x ph ) $=
( vy cab wtru wceq wsb wal cvv cv wcel wb dfcleq vextru tbt df-clab albii
bitr3i bitri weq df-v equid bitru abbii eqtri eqeq2i nfv sb8v 3bitr4i ) A
BDZEBDZFZABCGZCHZUJIFABHULCJZUJKZUOUKKZLZCHUNCUJUKMURUMCURUPUMUQUPBCNOACB
PRQSIUKUJIBBTZBDUKBUAUSEBUSBUBUCUDUEUFABCACUGUHUI $.
and this is one more reason to use https://us.metamath.org/mpeuni/bj-df-v.html as the definition (and this theorem does not require ax-8 nor df-clel either).
and this is one more reason to use us.metamath.org/mpeuni/bj-df-v.html
I definitely agree, I often use T.
for tautologies and it generally makes proofs simpler. I would also argue to use us.metamath.org/mpeuni/bj-df-nul.html for the same reason.
Edit a few proofs to reduce axiom usage, specifically:
All added DV conditions are with dummy variables.
Axiom usage here: https://github.com/metamath/set.mm/commit/781a0bb89850b0271ab3b6d5288a598a6f318f57