At first sight it would seem that the recently popular theory in neuroscience called Predictive Coding is merely catching up with insights long-known in the phenomenology of Husserl - such as the essential interplay for cognition of retentions and protentions and of a nested hierarchy of horizons - and also with the Merleau-Pontyian insistence on the functional integration of perception, proprioception and interioception. In this case however the apparent isomorphism between descriptions from the inside and from the outside mask a much deeper divergence. Even in its anti-transcendental formulation, phenomenology takes knowledge and freedom as paradigmatic, while the neuroscientific accounts, which can be unified under some such least-action principle as the minimisation of free energy actually tend to explain the internal hypotheses of freedom and knowledge as approximations which can in principle never attain full precision. This difference goes to the very heart of the meaning of subjective experience. If such a thing as qualia, say, or some other avatars of pure subjectivity, exist then surely their  minimal function is to underpin phenomenology. But if phenomenology is never what it thinks it is, this function may well be redundant.