With that rationale, what prevents me from just posting a stuff() function call as an answer to everything and claim that the implementation of that function is found in another part of the project? Because a single function containing nothing but a function call to another function is very valid in pretty much any programming language.

@Lundin I don’t follow. Your code needs to contain all the code it references otherwise it won’t be able to run. If you use stuff(), you must define stuff().

@Lundin this answer is only about the specific problem of allowing functions as answers. My answer deals with the requirements for the actual language - you have 2 "places" where stuff() could be defined - your answer code whose bytecount you count, and the language you're answering in (whose size you obviously ignore). One of those must define stuff() to do what you want it do, and if that's not your answer, it's the language, at which point you stop competing with other C answers.

My point is - your lambda won't run as-is, you need to wrap it in some other code not present. How is that different from calling a function not present?

@Lundin A function is just allowed as a form of an answer (as an alternative to a full program). e.g. otherwise you are restricted to only a single string I/O (stdout/stdin) which is often extremely limiting (e.g. you'd need to have a JSON parser equivalent to take, as an input, an array of arrays of arbitrary strings, which is just purely stupid). A discussion about what are valid "programs" should be another meta question, but is mostly unrelated to this one.

@Lundin Wrapping a function in other code is only an I/O bridge, not code that actually affects how the function would run and what it would return.