In the Weights measures your AI footprint

Thomas Dimson and Joey Flynn, both former employees of OpenAI, have launched a product called In the Weights. The tool purports to measure how well an artificial intelligence model is able to recall a specific person without relying on external tools like web search. Dimson and Flynn previously co-founded Global Illumination, a design startup that OpenAI acquired. Their website introduces a spin on the traditional vanity search—the act of searching for one’s own name—by looking at what is stored directly inside the AI models themselves. According to the In the Weights website, being in the weights indicates that a person’s existence was considered important during the creation of superhuman artificial intelligence.

To calculate this footprint, the tool supposedly queries 13 chatbots to assign a strength score. These models include Grok, Gemini, GPT, Claude, and Llama. The tool queries the models with a prompt asking for up to 10 results per query, including a short description and a confidence rating, before clustering the descriptions to generate a final score. For example, the author of the source article received a strength score of 641, placing them in the top 6% of names. Meanwhile, the leaderboard has shown other names at the top, with Macaulay Culkin reaching a strength score of 988, closely competing with Luciano Pavarotti.

The tool also exposes how models can generate false information, known as hallucinations. For instance, GPT-5.4 Mini described Anthony Ha as an “ambiguous name form that could refer to multiple people with the initials A.H.A.” This output highlights the limitations of relying on LLMs (Large Language Models) for factual biography. AI critic Anthony Moser dismissed the tool, calling it identical to asking 13 chatbots to describe oneself.

Despite the criticism, Dimson noted that the reception to the site has been intense. He explained that he and Flynn built the project to stimulate their creativity again after leaving OpenAI. Dimson argues that “Google vanity searches are the wrong objective in 2026 as more traffic moves to LLMs,” pointing out that many lives are recorded within the floating point numbers of AI systems. The weights he refers to are the numerical parameters that shape an AI model’s training and output, representing a shift in how personal data is archived and retrieved.

Why it matters

The tool highlights how AI models are replacing traditional web search as the canonical source of information. As more traffic moves to LLMs, the creators note that Google vanity searches are the wrong objective in 2026.