New AI Tool 'In the Weights' Reveals if You Exist Inside the Machine's Memory

Anyone who has Googled their own name recently knows the experience has changed. As traditional web search loses its monopoly on information, a new question has emerged: does artificial intelligence know who you are? [3] Former OpenAI employees Thomas Dimson and Joey Flynn have launched "In the Weights," a novel web tool that functions as an AI-centric vanity search. [1][4] The platform allows users to check if their identity and biography are permanently encoded into the neural networks of the world's leading AI models. [2][1][2][3][4]

The tool operates by simultaneously querying a diverse roster of frontier and open-source models, including OpenAI's GPT-5.5, Anthropic's Claude Opus 4.8, Google's Gemini, xAI's Grok, and Meta's Llama series. [4] Instead of relying on external web searches or browsing capabilities, the site forces the models to rely entirely on their internal memory. [2][3] It asks each system to identify a specific person and provide a short description, testing whether the individual's existence was deemed significant enough during the training process to be baked into the AI's core architecture. [3][6][2][3][4][6]

The project's name refers to the billions of numerical parameters—or "weights"—that form an artificial intelligence's brain. [5][6] When an AI model is trained on vast datasets, it adjusts these weights to encode relationships, facts, and language patterns. [6] If a model can recall a person accurately without looking them up on the internet, that individual is officially "in the weights." [1][2] Dimson notes that this metric serves as a unique mirror for the modern era, reflecting whose lives and legacies have been immortalized in the foundational technology of the future. [6][1][2][5][6]

To quantify this digital footprint, the platform aggregates the responses from the various models, clusters similar descriptions together, and assigns a unified "strength score." [3][4] The scoring system runs from zero up to a theoretical ceiling of 996. [4] According to the creators, the absolute highest scores are generally reserved for universally recognized historical figures like William Shakespeare and Wolfgang Amadeus Mozart, or massive contemporary pop stars like Taylor Swift. [2][4][2][3][4]

The platform's public leaderboard has already generated significant interest by revealing how current models prioritize modern celebrity. [5] As of late June 2026, actor Macaulay Culkin holds the top spot among living figures with a staggering strength score of 988. [1][5] He is followed closely by the late opera singer Luciano Pavarotti. [3][5] Meanwhile, a score of 641 is enough to place an individual in the top six percent of all searched names, highlighting just how exclusive the upper echelons of AI memory truly are. [3][5][1][3][5]

One of the most revealing insights from the tool involves the size of the AI models being queried. [4] The creators emphasize that appearing in smaller, open-source models is actually a stronger indicator of global relevance than showing up in massive frontier models. [2] Because compact systems like Meta's 1-billion-parameter Llama model must compress knowledge far more aggressively, they filter out niche information. [2][4] Making the cut in a tiny model means the training data prioritized that individual's existence over millions of others. [4][2][4]

However, the tool also exposes the persistent flaws and limitations of modern language models. [2] The aggregated results frequently highlight AI "hallucinations," where models confidently invent biographical details or attribute false achievements to real people. [3][6] The creators warn that typos in a search query can drastically drag down a strength score, and individuals with common names often receive muddled or inaccurate descriptions as the models struggle to disambiguate them from others with the same moniker. [2][4][2][3][4][6]

Beyond its entertainment value, "In the Weights" visualizes a critical transition in the digital economy. [4] For researchers, journalists, founders, and creators, professional reach is increasingly dependent on AI-mediated discoverability. [4] As internet traffic continues to shift away from traditional search engine indexes and toward conversational interfaces, a model's ability to accurately recall an individual is becoming the new baseline for relevance. [4][6][4][6]

Ultimately, the retro-styled platform answers a growing cultural curiosity about our relationship with superhuman technology. [3][4] It bridges the gap between complex machine learning architecture and everyday vanity, offering a tangible way to measure digital legacy. [5] Whether it remains a passing curiosity or evolves into a standard metric for digital footprint analysis, the tool proves that the desire to know if we exist inside the machine is universally compelling. [3][3][4][5]