4Fix a model M of GCH. Let I be uncountable in M , let P = I,2, and let G be P-generic over M . Let N = L(P())M [G] , the constructible hull in M [H] around P() M [H]. We argue N does not model choice, though since M [G] models ZFC, N must model ZF; this would imply Con(ZF) Con(ZF+AC). So, suppose N models AC, and let be an ordinal in N (necessarily, M as well, since forcing adds no ordinals) such that ( = |P()|)L(P()) in M [G]. Now take M which is a regular cardinal containing in M , let Q = Fn(, 2), and let H be Q-generic over M . By E1, E2, and E3, we know that for any ordinal (including ) in M , and any formula (x), ()L(P()) is true in M [G] if and only if 1 P ( ) if and only if 1 Q ( ) if and only if ()L(P()) is true in M [H]; thus, in M [H], ( = |P()|)L(P()) . But (absolutely, in every transitive model) so, by preservation of cardinals, is not in bijection with in M [H]. But = |P()| in M [H], since GCH holds in M , is regular in M , and we are forcing over Fn( , 2), so there is a bijection there. But by preservation of ( = |P()|)L(P()) , there is a bijection (constructible over P()) between and P(), violating the cardinal-hood of in M [H], a contradiction. It should be noted that N is a model of V = L(P()), by absoluteness of P and D. It is thus also a model of V = HOD(P()), since ZFC proves X(L(X) HOD(X)). So, our actual implication is Con(ZF) Con(ZF + AC + V = L(P()) + V = HOD(P())).