A list of the most important unsolved problems in VFX production should include the word "pipeline". Pipelines are a major source of efficiency, but they require ongoing development, and are seemingly never finished. They also resist change. Is it possible to automatically generate pipelines given desired inputs and outputs or design one "universal" pipeline that can do everything with no further modification ever needed? We present similarities between pipelines and programs, and recall metamathematical statements which show that the answer to the above question is negative. However, new approximate solutions may someday mitigate to a large extent this negative situation.