I tapped at my computer. I had spent so long searching for Meng, but I had never thought to search for grandmother. I tried all the tricks I had learned, and the results came up similarly blank. It was as if the world had accepted she was fading and was prepared to let her fully disappear.

For the first time since I connected with Louise, I dove directly into the EMBRS data. There’s a lot more now, not only social media posts. We’ve been building up its knowledge base as we prepare the program to better understand journal entries. Prof. Logan has assembled a moderation team of minimum-wage undergrad students. They don’t have access to the journal entries themselves, but they can approve connection topics—yes, connect mothers returning to the workforce; no, do not connect those scheming violent insurrectionists.

Prof. Logan’s latest pitch had been in San Francisco, and we’d recently received an influx of users from California, really the first to use EMBRS as a journal. Unfortunately, there had been a lot of false positives, so he had me work on ingesting local newspaper articles and historical archives and such. The data was meant to inform EMBRS about the region so it could better determine a unique connection point between users. I decided to do something I had never done before: I fed all of my previous journal entries into the system.

I made sure it wouldn’t reach out to random test users and tell them about me. But it would parse my entries and try to find connections for me among its data. I didn’t think it would actually surface anything, just that my journal entries would be realistic test data.

After running for eight minutes (I know, we need to optimize), EMBRS returned a single result, from the archives of one of the local newspapers.

The fire appears to have been a grease fire, though officials were unable to find an occupant of the house. The fire spread unusually quickly due to the amount of flammable material near the kitchen including paper, cardboard boxes, and, most peculiarly, pencils. Most of the wood of the pencils burned, leaving strips of lead at times, piles of ashes, others. Neighbors varied widely in their statements on who lived in the house. Some said it was unoccupied, while others insisted a large family lived there, and one even thought it was a restaurant.

I almost dismissed the article as another false positive. But then EMBRS showed my journal entry that matched. It was the conversation between me and grandmother, when she said she’d lost most of her pencils in a fire in California. I nearly slammed my laptop shut.