A few challenges that requires the program to output something at a specified length, and a bunch of answers well... uses built-in error messages that's generated by the interpreter rather than the program itself.
Say for example, this program outputs a total of 20 bytes (not counting trailing newline)
#!/bin/bash
@
with output
@: command not found
Of course this message will not appear in STDOUT, but STDERR instead (without redirection).
Should this count as a valid program that outputs something as long as 20 bytes?