Urgent.News

What's breaking now, across thousands of outlets.

AI

Marigold — a private ofrenda that catches the ghosts AI writes into our memories

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine What I Built For Día de los Muertos, families build an ofrenda : an altar with photos of the people they've lost, their favourite foods, marigolds to guide them home, and the stories that keep them close. Marigold helps you build one on your own computer: Add someone and drop in a few old photos.…

Marigold is an AI-powered app that allows users to create a personal ofrenda, a Día de los Muertos tradition, on their own computer. Users add photos of their loved ones, write memories, and the app generates a short remembrance using a local open-weight model, Gemma 3 4B via Ollama. The ofrenda is saved as a single HTML file that can be kept, printed, or shared.

To ensure the accuracy of the remembrance, Marigold incorporates a "ghost check" that flags any invented names, places, numbers, months, or weekdays. The model is programmed to only use the information provided by the user and what it sees in the photos. If the AI generates anything that doesn't align with the provided facts, it is highlighted and can be removed or added to the memories if it turns out to be true.

Marigold's developers used a local open-weight model, Gemma 3 4B, which reads images and text. The app is built using .NET 8 and Ollama, with no third-party dependencies, CDN scripts, or web fonts. The model was tested extensively, with 33 built-in self-tests for the ghost check, photo validation, storage, and the ofrenda page. The ofrenda page is a single file with embedded photos as data URIs, no scripts, no links, and no fonts to download, making it lightweight and privacy-friendly.

By building an ofrenda on their own computer, users can keep their memories private and have a personalized, respectful tribute to their loved ones. The AI's ghost check ensures that the remembrance stays true to the facts shared by the user, preventing any potential "ghosts" that might rewrite someone's life.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

AI’s Last Mile

Physical AI: From Newsline to Production As a young man, I never missed Newsline on NTA on Sunday nights. Frank Olize’s voice belonged to the weekend as surely as Sunday rice.

More from Saturday 31 October →