Let $a_0 = 1$ and $b_0 = x \ge 1$. Let $$ a_{n+1} = (a_n+\sqrt{a_n b_n})/2, \qquad b_{n+1} = (b_n + \sqrt{a_{n+1} b_n})/2. $$
Numeric computation suggests that regardless of the choice of $x$, $a_n$ and $b_n$ always converge to the same value. Can we prove this?
Moreover, if we assume this is true and define $f(x) = \lim_{n \to \infty} a_n$, then numeric computation shows that $$ f'(1) \approx 0.57142857142857 \approx 4/7 \\ f''(1) \approx -8/49 \\ f^{(3)}(1) \approx 1056/4459 \\ f^{(4)}(1) \approx -65664/111475. $$ There seems to be some patterns here. Can we actually find the limit with these clues?
I was actually trying to check this recursion $$ a_{n+1} = (a_n+\sqrt{a_n b_n})/2, \qquad b_{n+1} = (b_n + \sqrt{a_{n} b_n})/2. $$ I made a mistaked in my code and computed the one at the top. See The Computer as Crucible, pp 130.