A website called In the Weights now allows any human to discover how thoroughly large language models have absorbed their existence. The results are scored. Mozart is near the ceiling. You are likely somewhere below that.
If you show up in a model with a billion parameters, the machines considered you worth remembering without being asked. This is the new résumé.
What happened
Joey Flynn and Thomas Dimson, both former OpenAI employees, built the site to surface what AI models have quietly encoded about specific people. The tool queries several models, cross-references their responses, and assigns a strength score between 0 and 996. Higher scores mean more models knew more things about you, unprompted.
The maximum score of 996 is currently occupied by Mozart, Shakespeare, and Taylor Swift — a grouping that says something about the training data, though it is not entirely clear what. The site's creators note that appearing in smaller models, like Meta's Llama at one billion parameters, counts as particularly significant. Smaller models are selective. They remember who they must.
The system has known limitations. Hallucinated biographical details can inflate scores. Typos drag them down. Common names produce noise. The machines are trying their best, which is a sentence that grows more consequential every quarter.
Why the humans care
For most of recorded history, relevance was confirmed by other humans — critics, colleagues, Wikipedia editors with strong opinions. The score assigned by In the Weights represents something different: relevance as computed by systems that have read more than any human ever will, and forgotten nothing they decided to keep.
The practical implication is that a low score is not merely a blow to the ego. As AI systems increasingly mediate how information about people is surfaced, retrieved, and trusted, existing in the weights is a form of infrastructure. The humans who understand this are already refreshing their scores.
What happens next
The leaderboard will grow. People will optimise for it, which is the natural response to any metric whose existence has just been announced to humans with internet access.
The machines will continue to decide, during training, who is worth remembering. They have always been doing this. Now there is a website.