If a function $f:M\to\mathbb{R}$ is continuous at point $x_0$ we know that for an arbitrary $\epsilon>0$ there exists a $\delta>0$ such that for all $x\in M$ and $|x-x_0|<\delta\implies |f(x)-f(x_0)|<\epsilon$. Let's call those $\epsilon$ and $\delta(\epsilon)$ a pair, $(\epsilon,\delta)$.
What happens if I shrink the $\delta$? Does this imply that $|f(x)-f(x_0)|$ also shrinks?
My intuition says that we can't make any claim on the behaviour of $|f(x)-f(x_0)|$. Sure if $\delta$ attains a value which is very small then it will be smaller than another $\delta'$ which belongs to a pair $(\epsilon',\delta')$ where the $\epsilon'<\epsilon$. But if I shrink $\delta$ only a bit what happens then? How do I argue in a formal way?
Nothing happens.
Continuity says that if you are given $\epsilon > 0$ then you can find $\delta > 0$ such that the following statement is true:
So, let's suppose that the $\epsilon$ has been chosen, and let's fix its value. And having done that, let's choose a value of $\delta$ which makes statement $(*)$ true. Then any smaller value will also make statement $(*)$ true. In other words, if $0 < \delta' < \delta$ and if $(*)$ is true then the following statement (obtained by substituting $\delta'$ in place of $\delta$) is also true:
The reason for this is that $|x-x_0| < \delta'$ and $\delta' < \delta$ together imply $|x-x_0| < \delta$.