The pipeline drafts the walk between stops
A quiet stretch on the walking app and the API — nothing new to report there this time — but the research pipeline that feeds them both got a genuinely useful upgrade.
For historians: the first draft writes itself
Last week’s log introduced the approach story — the narrative beat for the walk toward a waypoint, not just the stop itself. Until now, writing one meant opening a waypoint and starting from a blank box. As of this week, the pipeline drafts it for you: a new stage runs after the main narrative pass and writes an approach_narrative for every waypoint of every trail, in trail order, so there’s always something to react to and edit rather than a blank page to fill.
The draft isn’t invented from nothing, either. A new landmark-research step walks the actual corridor between two waypoints — using real map data to find what a person would genuinely pass along that stretch — and hands those to the writer as source material. It’s the same principle as the rest of Suburb Historian’s drafting: specific and grounded, not generic scene-setting.
And because trails change — a waypoint gets reordered, a geocode gets corrected — an approach story can go stale without anyone noticing. The pipeline now tracks that: reorder a trail or fix a location, and every approach story touched by the change is flagged with a banner pointing straight at the “✨ AI rewrite” button, so nothing publishes describing a walk that no longer matches the trail. (hopefully that little AI Rewrite icon works)
Under the hood
A separate, longstanding annoyance with dragging a waypoint pin on the region map is fixed: the marker itself would snap and jitter mid-drag because its position and icon were being handed to the map library as new objects on every render, even when nothing about them had actually changed. Both are now cached and only replaced when they genuinely differ, so dragging a pin feels as smooth as it should. (This is separate from — and doesn’t touch — the geofence-circle lag fix from a few days earlier; that one was about the dashed radius circle catching up to a dragged pin, this one’s about the pin itself.)