If $n>m>0, n, m \in \Bbb{N}$, is $n^m-m^n$ positive or negative? (Explain with cases if needed.)
First, for finding patterns, I've put some small integers to $(n, m).$
\begin{align} (2, 1): \; & 2^1-1^2=1>0 \\ (3, 2): \; & 3^2-2^3=1>0 \\ (4, 3): \; & 4^3-3^4=-17<0 \\ (5, 4): \; & 5^4-4^5=-399<0 \\ \end{align}
...and gained nothing.
Also, I've found some trivial theorem about this, like:
Thm 1. $n^1-1^n>0$ for all $n>1$.
No need to prove it.
Thm 2. $n^2-2^n<0$ for all $n>4$.
Proof:
Let's draw a graph for this.
The red one is a graph of $y=x^2$, and the blue one is a graph of $y=2^x$.
As you can see, $y=2^x$ increases more exponentially then $y=x^2$.(Also a trivial thing without graph)
The intersection of two graphs in positive reals is $(2, 4), (4, 16)$.
Therefore, $n^2-2^n<0$ for all $n>4$.
I think there is a particular integer $N$ which satisfies, $n^N-N^n<0$ for $\forall n>N$.
But I can't do anything more...

In general $(n, m)>0 $ whenever $n<m$. $$ {} $$
The ONLY exceptions are that $( n, 1)> 0$ for $n>1$ and $(2, 3)<0 $
– WW1 Jan 23 '23 at 20:13