Abstract: Languages have ways of indicating whether the available evidence for an utterance is direct or indirect/inferred; what we often call ‘evidentiality’ (see e.g., Aikhenvald 2004). Cross-linguistically, evidential distinctions are mostly marked in the past or the present. When the future is used as an evidential it either loses its futurity (e.g., French, Italian, Greek, Dutch, German (Haegeman 1983; Comrie 1985; Palmer 2001; Kush 2011; Matthewson 2012; Giannakidou & Mari 2018; Ippolito & Farkas 2021; Frana & Menéndez-Benito 2019, 2023 a.o.)) or it is an indirect evidential (e.g., Muylaq’ Aymara, Sanuma (Borgman 1990, Coler 2014)). Therefore, there seems to be a typological gap: we do not observe any direct evidential that is also a future modal (e.g., Aikhenvald 2004, Dall'Aglio Hattnher 2013). Based on original fieldwork from Colloquial Jakartan Indonesian (Tsilia 2025), I will show that this is not a real typological gap. I will focus on ‘mau’ in Indonesian, which can be both a desire verb (meaning ‘want’) and a proximate future (meaning ‘about to’). I will argue that when it is a proximate future, it requires direct evidence for the proposition it takes. But, how can one have direct evidence of the future? I will argue that language has precisely a way of talking about what counts as ‘direct enough’ evidence of the future. The future usage of ‘mau’ in Indonesian requires evidence for a proposition that directly leads to the proposition about the future. This introduces a new category of evidentials, namely direct inferential evidentials, which involve such a small reasoning step that for the purposes of linguistic marking it does not count as reasoning. Formally, building on previous work on epistemic ‘must’ as an indirect evidential and as a marker of cognitive complexity (von Fintel & Gillies 2010, 2021, Wurm 2023), I model what counts as ‘direct enough’ evidence of the future, starting from the set of propositions that are direct in the context and inductively defining a nested sequence of sets of propositions. Finally, I will present some ongoing work on negated proximate futures cross-linguistically, showing that an unexpected reading arises, similar to a desire reading, even when the proximate future is not ambiguous between ‘want’ and ‘will’ morphologically.