> some types of diets seem to help with autism

Sorry, the actual evidence for this is essential nil. The anecdotes are all explainable by confirmation bias. https://skeptics.stackexchange.com/a/43957/82