Many semanticists have claimed that must's meaning is weaker than epistemic necessity, a claim that von Fintel & Gillies (2010) dub "The Mantr". Recently von Fintel & Gillies have argued in an influential paper that the Mantra is false, and that the intuitions that have driven it can be accounted for by appealing to evidential meaning. I show that von Fintel & Gillies do not provide a compelling argument against the Mantra, and that their theory of evidential meaning, while promising in certain respects, also has serious empirical and conceptual problems. In addition, a variety of corpus examples indicate that speakers who assert must p are not always maximally confident in the truth of p. As an alternative, I reimplement on Fintel & Gillies' theory of indirect evidentiality in a probabilistic, Mantra-compatible framework. Ultimately, both sides of the debate are partly right: must is weak in several respects, but it also encodes an indirect evidential meaning.