I'd normally call that a low-effort, troll comment. But, thinking on it, you may have a great metaphor.

They keep promising great performance out of models whose key ingredient (parameters) they are diluting. Many seem to be in a competition saying they're getting smaller and higher performance at the same time. Then, the homeopathic models don't perform as well as real models when independently tested. Again, spot on.