Here's one that's amazing, more for its history than anything else:

tr -cs A-Za-z '\n' |
tr A-Z a-z |
sort |
uniq -c |
sort -rn |
sed ${1}q

(source and details: leancrew.com/all-this/2011/12/more-shell-less-egg)

This is Doug McIlroy's six-command response to Donald Knuth's "10 pages of pascal" solution to the challenge: Read a file of text, determine the n most frequently used words, and print out a sorted list of those words along with their frequencies.

It was a cheeky response because Knuth was actually demonstrating 'literate programming', which was intentionally verbose; and even McIlroy admitted Knuth's solution was more refined/powerful/extensible. But it was a great way to show the power of unix pipelines.

Which goes back to my original thought for this thread - it's not the individual commands that are amazing, it's how those simple pieces combine that breaks your head ;)

I don't know that it's "amazing" but type commandname is really useful; reveals a lot of information and works in cases where which and declare would fail.

Simple example:

$ which ls
/bin/ls

$ declare -f ls

$ type ls
ls is aliased to `ls -la'

which gives an accurate result but does not tell you there is an alias; declare wouldn't tell you anything; but type alerts you to the alias.

Makes it useful for beginners to learn; but also remains useful to more experienced devs. eg. in scripts trying to determine what's available on a server or workstation (I once needed to detect NVM and didn't realise it was a function, so which didn't help; but type did).

oh and another one

strace -p <process id>

:(){ :|: & };:

.... but please read what it does first :)

Well..

apt-get upgrade

Easy and works.

rm -rf

Such ease of use; such amazing power.

(Disclaimer: don't actually run that)