There is a lot of talk about how you should be building Domain Specific Languages to solve your issues. The argument is that you are naturally better off with a language specifically tailored to your problem domain's needs. I think there are a couple of serious flaws with this theory:
For many domains, their models are thrown away numerous times as new parts of the domain are discovered.
For me, my first introduction to domain specific languages was trying to decipher a math book my uncle wrote on tensor calculus.
Yesterday I spent an hour debugging a regex and ultimately coming up with a 4 character fix. So yes, Succintness is Power and Regular Expressions are a great example of
RegEx is a clever way to model FSMs but as a tool it leaves me with splinters.
If you want power, you can always stick your finger in a light socket...