The compose dock carries a pre-send "⏳" chip that forecasts the causal chain: how far, in years, your chosen moment lands from your nearest foothold (the objective's era, or a change you've already made). Land close and the message hits at full force; land in the deep past with nothing bridging it and the swing decays.

The chip only ever surfaced a caution. Its best tier — at-hinge, a foothold landing at full force — returned null, so the chip vanished. Doing the right thing showed nothing, and null carried two meanings at once: "you nailed the chain" and "the chain doesn't apply here" (an anchorless objective, or no footholds landed yet). A player on a perfect chain and a player off the grid saw the same blank. In playtesting the system read as "invisible".

This change surfaces at-hinge as a positive chip and lets null mean one thing. The blast radius is small: the chip's data already existed in the store; the fix is two lines of view logic plus tuned copy, pinned by a new end-to-end spec. The four files below.

The view did the hiding, in two spots in pages/play.vue. The template now colors the chip by tier — at-hinge rv-ok (a win), diffuse rv-warn (a caution), the two middle tiers neutral — and the chainRead computed drops the tier === 'at-hinge' arm of its early-return, so every engaged tier reaches the chip. The only state that still returns null is the genuinely-absent one.

That null now means exactly one thing, because it is a straight pass-through of the store getter. causalChainRead builds the footholds (the objective's anchor plus every dated change) and asks chainStatus, which returns null only for an ungrounded contact or an anchorless objective with nothing landed yet — never for at-hinge.