Representational or semantic content plays an essential role in classical accounts of computation and cognition. Both cognitive science and computer science share philosophical foundations that endorse such a view. This has led quite naturally to a situation where computational theories of mind, theories which unite computation with cognition, have almost unanimously assumed that mental computation must also involve semantic content (see Pitt 2013). I will argue that mechanistic computationalism, as presented by Piccinini, can offer a genuinely non-semantic account of mental computation. Furthermore, I claim that if this account is correct it would provide important constraints on any theory of mental content. This is in contrast with Piccinini's position, which is that mechanistic computationalism should simply remain neutral on which theory of semantic content, if any, we should prefer.