Negate the definition of boundedness: Β¬(βMΒ βxβ(0,1]Β (β£f(x)β£β€M)).
Pushing the negation inward gives βMΒ Β¬(βxβ(0,1]Β (β£f(x)β£β€M)), then βMΒ βxβ(0,1]Β Β¬(β£f(x)β£β€M), i.e. βMΒ βxβ(0,1]Β (β£f(x)β£>M).
In words: "whatever bound M is proposed, there is a point x exceeding it" β that is the definition of unboundedness.
Now apply it to f(x)=1/x (xβ(0,1]). Let an arbitrary MβR be given. Set x=β£Mβ£+21β; the denominator is positive so x>0, and β£Mβ£+2β₯2 gives xβ€21β<1 β hence xβ(0,1].
Then β£f(x)β£=f(x)=β£Mβ£+2, and since β£Mβ£β₯M we have β£Mβ£+2>M. So β£f(x)β£>M holds at this x.
Since M was arbitrary, for every M there is an xβ(0,1] with β£f(x)β£>M β i.e. βMΒ βxβ(0,1]Β (β£f(x)β£>M) holds, so f is unbounded on (0,1]. The key is the quantifier order: because M is given first (βM), the witness x may be chosen afterwards (βx), which is why the M-dependent choice x=1/(β£Mβ£+2) is legitimate.