Musona ships offscreen lives, guardrails

This is a meaningful product-tuning update, not just a feature list. The interesting part is that Musona is iterating on the hard parts of companion AI: keeping the world coherent, preserving persona boundaries, and avoiding memory that feels omniscient.

• –surfaced_at is the right kind of mechanism for offscreen memory, because it treats unseen events as lower-confidence rather than permanently authoritative
• –Tightening the "user's stuff isn't your stuff" guardrail suggests the team is fighting persona bleed with prompt-level constraints, which is a real failure mode in companion products
• –The unresolved assertion-grounding problem is the biggest remaining risk: if the app can't resist false user claims, the illusion of continuity can collapse fast
• –Android availability matters here because these systems need repeated, real-world usage to prove whether the simulated life actually feels coherent over time
• –No Product Hunt presence surfaced in search, so this reads more like an indie product iteration than a launch-page moment