future app · harnesses Maps history idea AI-enhanced
stamp
Your Apple/Google Maps history, but only the places you went back to. stamp surfaces your real haunts — the bench, the bar, the corner — as tiny map-card kliplets.
"Places that earned a return."
The hook
Anyone who's ever opened Maps and thought "wait, I've been here before." stamp reads location history, finds the repeat-visit pattern, and turns each haunt into a klippable map tile.
The klip-feed
| Replaces | The Foursquare check-in reflex (RIP) |
| Klip payload | 1×1 map tile + Apple/Google Maps deep-link |
| Profile signal | A personal map dotted with "places that earned a return" |
AI layer
| What it does | A clustering model resolves GPS noise into real "places" (50–100m bucket with semantic dedup) and ranks them by haunt-score (frequency × recency × dwell). An LLM writes the one-line caption for each haunt ("Tuesday-after-work bar"). |
| Input | Raw location history + venue database + visit timestamps |
| Output | Resolved venue + haunt-score + one-line caption |
| Why with > without | Without AI, GPS coords just point at parking lots and intersections. With AI, the haunt is the bar, the bench, the bookstore — and it has a name. |
Open questions
- iOS location history — Apple's Significant Locations is on-device but not exposed via public API. May need user-driven import or rely on Google Maps Timeline export.
- Privacy — repeat-visit data is sensitive. Default to "klip a haunt manually" rather than auto-surfacing everything.
- Overlap watch — pin (first-times) + bench (sit-spots) + stamp (return-visits) form a triad. Keep distinct or unify under one places app?
Status
idea Filed 2026-05-15 from brainstorm session.
Lino Labs · future-apps · stamp