The era of the vanity search, once a reliable digital staple for the curious and the egotistical, is undergoing a profound transformation. A quick query for your own name on Google no longer yields the familiar, comprehensive dossier of links and social profiles it once did. This shift is not merely a symptom of the search engine’s evolving algorithms; it reflects a deeper, more tectonic change in where authority and information now reside. An increasing number of people are forming their first impressions of the world, and of each other, not from indexed web pages, but from the conversational outputs of large language models (LLMs). In response to this new reality, Thomas Dimson and Joey Flynn have launched In the Weights, a novel website that has turned the concept of the vanity search on its head, creating a kind of digital immortality leaderboard by measuring how well a person’s identity is encoded within the very architecture of artificial intelligence.
What Does It Mean to Be “In the Weights”?
The term “weights” refers to the numerical parameters that are the foundational building blocks of any AI model. These are the coefficients that, during the training process, are adjusted to shape a model’s ability to recognize patterns, generate text, and make predictions. The site’s creators, who met at OpenAI after the acquisition of their design studio Global Illumination, describe the concept with a striking philosophical bent. “Being in the weights means your existence was deemed important in the process of creating superhuman artificial intelligence,” the website’s about page states. In essence, it is a metric for how significant your life and work are considered by the data and training processes that have produced the most advanced AI systems.
To determine this, In the Weights does not simply ask a single search engine. It conducts a sweeping, multi-model audit. The system queries a wide array of leading LLMs—including Grok, Gemini, multiple versions of GPT, Claude, Llama, and several lesser-known models—with a specific prompt: “Who is [name]? Give up to 10 results, each with a short description and confidence.” It then processes the responses, clustering similar descriptions together and assigning a composite “strength score.” This score is not a measure of fame in the human sense, but a reflection of how consistently and confidently a model’s training data “recalls” an individual’s identity without relying on the crutch of a live web search.
Why the Creator Abandoned the Traditional Google Search
The impetus for the project, as Dimson explained to TechCrunch, was born from a shared feeling of creative restlessness after leaving OpenAI and a deep-seated conviction that “Google vanity searches are the wrong objective in 2026.” The traditional web search has lost its canonical status, with users now turning to chatbots for information. The idea that a person’s life story is now “encoded somehow in a bunch of floating point numbers inside the AI brain” presented a compelling, and slightly unnerving, challenge. The final direction of the site was “sealed” by a tongue-in-cheek blog post that riffed on the concept of AI weights and drew inspiration from Terry Bisson’s classic short story, “They’re Made Out of Meat.”
A New Kind of Leaderboard for a New Kind of Legacy
The results of this AI vanity search, as they are displayed on the site’s leaderboard, are both fascinating and deeply provocative. The top slot, which has been shifting dynamically since the site’s launch, is currently held by “Home Alone” star Macaulay Culkin, who sits neck-and-neck with opera legend Luciano Pavarotti with a strength score of 988. This humble tech blogger, for example, received a score of 641, placing them in the top 6% of names—a respectable showing that was quickly dwarfed by the higher scores of several TechCrunch colleagues. The score itself is a composite that reflects the number of models that can identify the person, the confidence of those identifications, and the consistency of the descriptions across different architectures.
The system also provides a granular breakdown of which specific models returned which answers. For any given name, a user can see the “hallucinations” that the models produce. For instance, GPT-5.4 Mini might return a response that is an “ambiguous name form,” suggesting a name could refer to multiple people. This feature, while a side effect of the system’s core function, provides an intriguing window into the state of AI training data and its biases.
The Mechanics of the Search and the Future of AI Recall
Dimson and Flynn are not simply satisfied with creating a static leaderboard. They have a roadmap for deeper analysis. Dimson has outlined plans to investigate why different models within the same series, like different versions of GPT, return different results for the same person. This will help to expose the inherent biases that different training datasets and model architectures have towards specific types of people. A particularly compelling stated goal is to identify which people “should have a Wikipedia article but don’t,” using the model’s “knowledge” as a diagnostic tool for the gaps in our own collective record-keeping.
While the concept has been met with a mixture of intrigue and skepticism—with AI critic Anthony Moser scoffing that the site is “literally the same as asking 13 chatbots to tell you about yourself”—the creators have acknowledged that the reception has been “insane.” The site’s appeal, they believe, rests on a fundamental human desire to see if one’s life and work have been deemed important enough to “live forever in the super intelligence.” The comparison factor, which allows users to see how they rank against friends and public figures, only adds to the addictive nature of the tool.
In the Weights represents a significant departure from the traditional search engine. It is not a tool for finding information; it is a tool for measuring the extent to which that information has been absorbed and internalized by the most powerful, and most opaque, intelligence we have ever built. The site’s charming, Nintendo-inspired retro design belies the serious, almost existential question it poses: in a world where knowledge is increasingly defined by what an AI can recall, what does it truly mean to be remembered?
