SANKOFA
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote
SANKOFA is an AI agent that builds a time machine for any city, and tells you what is proven and what is imagined. The problem: the visual history of cities, and of African cities above all, is scattered across distant archives in Paris, Washington or Amsterdam. People in Porto-Novo or on Gorée Island almost never see what their street looked like a century ago. Meanwhile, generative AI can "age" any photo in one click, but it invents silently: a beautiful fake image of the past damages memory instead of serving it. What SANKOFA does: 1. Time machine for places. Give it a place and a year. A seven-step agent searches open archives (Wikimedia Commons, Europeana, by keyword and GPS), checks each image with Muse Spark, segments today's view with SAM 3.1, finds where the period photographer stood and aligns the real archive onto today's view. Every area gets a verdict: documented, unchanged or hypothesis. Muse Image only fills areas no archive covers, always labelled. The agent tests itself on a hidden archive to measure its reliability, and writes a cited story with no invented detail. 2. Country chronicles. It rebuilds the history of a country from its sources only: 786 events quoted word for word, cross-checked against a second source written by other authors, with the gaps shown instead of filled. 3. Documentaries. Every place and country becomes a short documentary: real archive photos, quotes on screen, public-domain music, full credits. Coverage: 16 places in 15 countries (Benin, Senegal, Mali, Ghana, Egypt, Morocco, South Africa, France, Spain, UK, Italy, Turkey, India, USA, Brazil), 350 openly licensed archive documents. Real archive overlays for Notre-Dame 1912, the Colosseum around 1880 and the Taj Mahal. Why it matters: heritage, education, tourism and restitution debates all need images people can trust. SANKOFA starts in Porto-Novo, my home region, and works for any city on Earth. Live showcase: https://sankofa-two-topaz.vercel.app