Why this exists.

More cold cases are being solved by the public now than ever before. People with patience, databases, and a reason to care have identified victims and named offenders that institutions alone could not. But that only works when the underlying facts are sound. The longer a case sits, the more a single misreported detail (a wrong date, a transposed digit, a guess repeated until it hardens into fact) gets treated as settled truth. Misinformation doesn't just mislead; it convinces someone holding a genuinely useful piece of information that it couldn't possibly matter. That's why Latent only ever shows what the investigating agency itself reported. No news clippings, no aggregators, no secondhand retellings. The primary record, or nothing.

And some of these cases were never going to be solved one jurisdiction at a time. We know there are offenders who crossed state lines and left behind victims we still can't name. Today, a researcher, a journalist, or a family member who suspects a connection has almost no way to ask the obvious question: this person was in this place, in these years; are there unsolved cases here that fit? Answering it means manually trawling dozens of incompatible agency websites, when the data is online at all. Latent is built to make that question answerable. Every connection it surfaces is a lead for a person to examine, never an accusation, never a name attached to a crime. The tool points; people decide.

These victims matter, and they don't stop mattering when the people who loved them are gone. As families age and pass on, a case can lose the one person who kept it in the conversation, who made sure it was still being worked, still remembered, still owed justice. Part of what this corpus does is simply refuse to let them disappear.

I'm Courtney Hammond. Latent exists because the resource it provides did not, and because the need was too important to leave unmet. For years I had watched cold-case data sit scattered, inconsistently structured, and incomplete, in a way that quietly fails victims, their families, and the agencies that rely on the public's help. What I lacked, until recently, were the technical skills to address it; those came with completing my bachelor's in computer science and beginning my master's. I can't think of a more worthwhile place to put them.

NamUs, which focuses on the unidentified, does essential work, but homicide victims who are named deserve the same standard of access and care, and they have not received it. The records that do exist are spread in scattered, incomplete fragments across individual agency sites. I won't pretend to know exactly why; the reasons are likely some combination of limited funding, lost case files, and institutional caution. Virginia has shown that it can be done well, yet almost no other state has followed. True-crime media helps keep these cases visible, and that matters, but it runs into the same walls anyone does: thin public records, older reporting of uncertain accuracy, and no way to see across cases, because until now no multi-state corpus existed. So I'm building one.

The project is independent and self-funded: no advertising, no investors, nothing for sale. Extraction runs entirely on my own hardware, no case data is handed to an outside company, and the models are never permitted to infer attributes such as a victim's race. The how of all of that lives on the Methodology page.