Animal tracking data mostly lives in research portals like Movebank, and the public sees it as an occasional news map or a conservation charity's "adopt an animal" pack with a certificate and a cuddly toy. Neither is a story. Migrant Postcards turns one real animal's track into a serialized, place-by-place journey: every stopover becomes a postcard in the animal's voice, with true facts about where it is and what it is doing, arriving roughly when the animal does.
The data it runs on
Source: public studies on Movebank (the Max Planck Institute of Animal Behavior's open tracking database), where researchers choose to share GPS tracks of storks, terns, whales and more, under each study's licence.
Access: the Movebank REST API pulls new locations for studies whose owners allow public download; we ask permission study by study and credit the team on every card.
Limits: stork GPS tags send data over mobile networks, so there are gaps over the Sahara and the sea. Whale satellite tags give a few fixes a day, each accurate to a few kilometres, and fall off after months. Small birds like terns carry light loggers that are read a year later, so their stories are replays. Many studies delay public data on purpose to protect the animals.
Fallback: when no live track is shared, the app replays a past season of a real tracked animal and says so clearly on every card.
1. The mission
Make millions of people care about migration by letting them follow one animal they love. When you have had a postcard from a stork in Mali, the wetland she depends on is no longer abstract.
2. Why people can't resist it
A pen pal with wings. First-person postcards make an animal a character you root for.
Real stakes, real time. Will she cross the Sahara? The gap in the signal is the cliffhanger.
Collecting. Every stop is a stamp. Three animals, two hemispheres, one album.
3. How it works for real
Ingest: nightly pull from Movebank for opted-in studies, cleaned with a speed filter to drop bad fixes.
Detect stops: plain code finds stopovers (more than 12 hours within 15 km) and daily distances.
Look up the place: reverse geocoding plus a fact table from Wikidata and species accounts: wetland names, habitat, weather.
Write: a frontier LLM (Claude or GPT via OpenRouter) writes the postcard in the animal's voice, using only facts from the table, and a fast model (Gemini Flash) checks each claim against it.
Paint: an image model (Flux or Gemini image) paints the front in a consistent vintage postcard style, from the place and habitat.
4. Cost per use
Free data; one image and text per postcard. About $0.03 for the painted front and $0.005 for the text, shared by every follower of that animal, so it is close to zero per user.
Free: follow one animal. Flock ($3 a month): follow every animal, printed postcards mailed for real (at cost), and 20% of revenue funds the tracking teams.
5. Growth loop
Classrooms follow one animal together; each child shares the postcard home. Families along the route (Tarifa, Mopti, Gorgona) get "she is passing over your town today" alerts and join. Research teams promote their own animal because it funds them.
6. Risks and guardrails
Animal safety: positions are delayed at least 3 days and blurred to 20 km for vulnerable species, so nobody can find a nest or a calf.
Honesty: the voice is playful, the facts are checked. Replays and gaps are labelled.
Death: animals die on migration. When a tag stops for good, the app tells the truth gently and explains what the scientists learned.
Credit and consent: only studies whose owners opt in, with their names on every card.
7. MVP build order
One stork study with permission, nightly pull, stop detection.
Map, daily distance chart, and the stop list.
Postcard writer with a fact table and the claim checker.