In the Weights: A New Tool to Measure How Well AI 'Remembers' You
In the Weights is a new website that evaluates how well various AI models can recall a person without web search, assigning a strength score.
Two former OpenAI employees, Thomas Dimson and Joey Flynn, have created a website called In the Weights that measures how well different AI models remember a specific person without using external tools like web search. The name refers to the weights – numerical parameters that shape an AI model's training and output.
In the Weights queries multiple models, including Grok, Gemini, various GPT versions, Claude, and Llama, as well as lesser-known models, with a question like: "Who is <name>? Give up to 10 results, each with a short description and confidence." It then clusters similar descriptions and assigns a "strength score." For example, TechCrunch journalist Anthony Ha received a score of 641, placing him in the top 6%. Currently, actor Macaulay Culkin leads with a score of 988, followed by opera singer Luciano Pavarotti.
The site also shows which models returned answers for a given name and highlights potential hallucinations – for instance, GPT-5.4 Mini considers Anthony Ha an "ambiguous name form that could refer to multiple people with the initials A.H.A."
Dimson explained that the idea came from thinking how "Google vanity searches are the wrong objective in 2026 as more traffic moves to LLMs" and that "so many lives are encoded somehow in a bunch of floating point numbers inside the AI brain." The direction was sealed by a tongue-in-cheek blog post riffing on AI weights and Terry Bisson's classic short story "They're Made Out of Meat."
Reception has been "insane so far," Dimson said, noting they thought it would be a mild curiosity but it seems to have struck a nerve about wanting to see if you live forever in the superintelligence. The site also features a cute, Nintendo-inspired retro design. Dimson plans to explore why different models in the same series return different results, which models are biased towards different types of people, and which people "should have a Wikipedia article but don't."
