Geospot
Projects PublishedFebruary 2026 Views9.4M from San Francisco, San Jose, New York City Geospot was an app that used image embeddings to guess where you were in the world from a single image and no GPS. It used GeoClIP before I trained my own embedding model with consented user data and data I found online. Geospot Infinity was an online RL game that used the same embedding model to learn to geog
Projects PublishedFebruary 2026 Views9.4M from San Francisco, San Jose, New York City Geospot was an app that used image embeddings to guess where you were in the world from a single image and no GPS. It used GeoClIP before I trained my own embedding model with consented user data and data I found online. Geospot Infinity was an online RL game that used the same embedding model to learn to geoguess from user feedback. As of August 2026 I'm working on gathering a corpus of a billion street view images and training the best geoguessing model in the world. Please reach out if you're…
related reading
- How Does A Blind Model See The Earth? — LessWronglesswrong.com
- segment-geospatialsamgeo.gishub.org
- LLM Visualizationbbycroft.net
- Embeddings: What they are and why they mattersimonwillison.net
- Blueprint-Bench: Testing spatial intelligence in AI models | Andon Labsandonlabs.com
- justinzwu.comjustinzwu.com
- How Does A Blind Model See The Earth? - by henryoutsidetext.substack.com
- Land Lineslines.chromeexperiments.com
- Fulltext Geocoding Versus Spatial Metadata for Large Text Archives: Towards a Geographically Enriched Wikipediadlib.org
- Yufeng Zhaoyufengzhao.com
- GitHub - ansonyuu/matchmaking: Embedding space of names clustered based on their interests using the sentence-transformers all-MiniLM-L6-v2 modelgithub.com
- Projects | devonzuegel.comdevonzuegel.com