I know that $-7^2 = -49$
Therefore $\sqrt{-49} = -7$
Because $\sqrt{-1} = i$ we can then expand it to $\sqrt{-49} = -7 = 7i$
And therefore $-7 = 7i$, divide both sides by 7 and you get
$-1 = i$
And I know that is not true, because $i = \sqrt{-1}$. What is wrong with my proof?
At risk of sounding tart, it's easier to say what's right with the proof:
$-7^{2} = -49$ is true (because with normal precedence of operations "$-7^{2}$" means "$-(7^{2})$".
$\sqrt{-1} = i$ is fine as notation, but will reliably start an argument in some quarters on Math.SE. (What vexes people is that $\sqrt{\ }$ is a function in the sense understood by The General Public only if the radicand is a non-negative real number, and $-1$ is not a non-negative real number.)
What's wrong:
$-(7^{2}) \neq (-7)^{2}$, so "Therefore $\sqrt{-49} = -7$" doesn't follow.
On using $\sqrt{-1} = i$ to expand $\sqrt{-49}$, two falsehoods get daisy-chained, $\sqrt{-49} = -7$ and $-7 = 7i$. Ironically, the combined effect, "$\sqrt{-49} = 7i$" is correct in the sense that $(7i)^{2} = -49$. This shows that in mathematics, two wrongs (sometimes) do make a right.
If we include errors of style, the introduction of $7$ into the proceedings does nothing to clarify. The entire proof could have been reduced to one erroneous step:
And then the sign error would have been apparent: $-1 = -1^{2} = -(1^{2}) \neq (-1)^{2} = 1$.
The real moral is the same as with debugging code: Always strive to reduce a contradiction to the smallest and/or simplest problematic situation possible.