Introduction
In most interpreted languages, there is an operation that will run code generated at runtime. It generally requires alphanumeric characters to refer to this operation. Write a program, using as few alphanumeric characters as possible, that is able to download and execute any string in the language you're working with, using as few alphanumeric characters as possible.
Informal bonus points for commonly used languages where this is a very difficult task, such as python.
Challenge
Inputs: a url to code in your language of choice, via either stdin or argv.
Outputs: output of executing the downloaded code.
Your program must generate the string to execute and call the eval/exec function/operator without reading from any existing files (excluding the file readings done by your interpreter of choice on startup and reading of standard library files by your language's import mechanism).
Winning entries must be written using only printable ascii characters. Compiled languages are allowed, if you can find a way to invoke the same compiler at runtime with the same settings.
Submissions will be scored, per-language, based on who uses the fewest executable alphanumeric characters. Non-alphanumeric characters do not count towards your score, nor do comments - try to make your non-alphanumeric code readable.
Input and Output example
Contents of my example target gist for python:
print("run by no-alphanum code exec")
example run:
$ python noalpha.py https://gist.githubusercontent.com/lahwran/77e94ef11493d4f36c726efe030f58e4/raw/5f7a0483a9fe7928ae2cf3f786d6d35f9d7870b9/example.py
run by no-alphanum code exec
$
printable ascii
, does that include newlines and tabs? \$\endgroup\$exec
oreval
? How much of the language needs to be interpreted? What if our language doesn't have a builtin function to execute code? \$\endgroup\$