@Dusk is mispriced because the real privacy boundary isn’t Phoenix itself, it’s the Phoenix↔Moonlight conversion seam. If you can observe when value crosses models, in what sizes, and who tends to be on the other side, you get a durable fingerprint. System reason: conversions emit a sparse but high-signal event stream (timestamps, amount bins, and counterparty reuse) that attackers can treat like a join key between the shielded and transparent worlds. Regulated actors also behave predictably for reporting and settlement, so round-lot sizes and time-of-day cadence become a second fingerprint that compounds linkability. In a dual-model chain the anonymity set does not compound smoothly; it resets at the seam, so one sloppy conversion can leak more than months of private transfers. This forces a trade: either accept worse UX and composability via fixed-size or batched conversions, or accept privacy that fails exactly where regulated users must touch the system. Implication: price $DUSK as unproven privacy until on-chain data shows sustained two-way Phoenix↔Moonlight flow with no measurable clustering signal across multiple epochs with no stable amount or timing bands. #dusk