I think the third condition is false conditional on Schanuel’s conjecture. For example, we can use Schanuel’s conjecture three times to rule out the possibility that $e$ is the solution to $2^{2^u}-u=93$. I hope and expect that this (or some simplification) provides a template for ruling out all possible existential definitions of $e$. If the above equation holds with $u=e$, then consider the four pairs of expressions: $$w=\ln 2, \ e^w=2\\ x=1, \ e^x=e\\ y=e \ln 2, \ e^y=2^e\\ z=(2^e) \ln 2,\ e^z=2^{2^e}$$ We have five independent algebraic relationships among them: $$e^z - e^x = 93\\ e^w = 2\\ x = 1\\ y = w e^x\\ z = w e^y$$ By Schanuel’s conjecture (in a [version][1] that looks natural constructively, and can be considered a contrapositive of the usual [statement][2]) there must be a rational linear dependence among the first four variables: $$a w + b x + c y + d z = 0$$ So the three pairs of expressions with $w,x,y$ have four independent algebraic relationships between them, where the first comes from dividing the last equation by $\ln 2$: $$a + b/w + c e^x + d e^y = 0\\ e^w = 2\\ x = 1\\ y = w e^x$$ By Schanuel’s conjecture there must be a rational linear dependence: $$p w + q x + r y = 0$$ So the two pairs of expressions with $w,x$ have three independent algebraic relationships, where the first comes from dividing the last equation by $\ln 2$: $$p + q/w + r e^x = 0\\ e^w = 2\\ x = 1$$ By Schanuel’s conjecture there must be a rational linear dependence $$s w + t x = 0$$ This would mean $\ln 2$ is rational, which is impossible, so the original equation does not hold for $u=e$. [1]: https://mathoverflow.net/a/163965/44143 [2]: https://en.m.wikipedia.org/wiki/Schanuel's_conjecture