What we let go.

Halfway through v1 we drew a line. Anything that wasn't reliable enough to play sixteen videos at once on a five-year-old Mac without getting dramatic about it was either going to ship boring and right, or be sent back for retooling.

A short list of things we sent back:

• Lava Lamp mode — beautiful on a fresh MacBook Pro, theatrical on anything older. Coming back when the shader scheduler is happy across the board.
• SmartFlex Grid — the brand carried baggage; the split/merge logic survived as Grid Designer.
• Mood matching and Mood board export — the clustering model felt chaotic-neutral. Off the matrix until it earns its return.
• Reaction camera — the feature flag is off in v1 by choice, not by accident.

05. Smart cell display, in particular

We used to have six framing modes: Fill to Face, Fill to Torso, Fill to Torso Only, Fill to Person, Zoom to Body, Off. They were correct and impossible to choose between. The 2026-05 cleanup collapsed them into one Smart mode with a per-cell decision tree, a tightness slider, and a Top/Center/Bottom anchor.

The Smart mode does the work the user used to do — read the cell aspect, look for a salient subject, decide whether a center crop is safe, fall back to a clean zoom-out if not. The result is that the cell stays composed when the playhead moves, and you stopped having a settings spreadsheet to manage.

08. The whole thesis, in one line

We'd rather ship boring and right than dramatic and intermittent. v1 is what's reliable. v1.1 brought back the things that earned their entrance music. The roadmap is what's still in the wings.