In the early 2000s, the ultimate indicator of professional status and cultural relevance was simple: "Googling yourself." To appear on the first page of a Google search was to exist. However, as of 2026, the landscape of information retrieval has shifted seismically. The era of the blue-link search engine, while not dead, is increasingly being superseded by the generative capabilities of Large Language Models (LLMs). As traffic flows away from traditional search indices and toward chatbots, a new, existential question has emerged: If you aren’t "in the weights"—the numerical parameters of a neural network—do you really exist in the modern digital consciousness?
Enter In the Weights, a new platform developed by Thomas Dimson and Joey Flynn. The site offers a provocative, albeit playful, diagnostic tool: it measures how effectively a wide array of AI models can recall a person without relying on real-time web search tools. By querying models like GPT-4, Claude, Gemini, and Grok, the platform seeks to determine if an individual’s identity has been truly "encoded" into the silicon brain of the next generation of super-intelligence.
The Chronology of an AI Curiosity
The project’s genesis is rooted in the post-OpenAI careers of its creators. Dimson and Flynn, both former employees at OpenAI—having joined the organization following the acquisition of their design startup, Global Illumination—found themselves at a crossroads. Seeking a creative outlet that would challenge their understanding of how modern LLMs process information, they began discussing the limitations of traditional SEO.
"Google vanity searches are the wrong objective in 2026 as more traffic moves to LLMs," Dimson noted in correspondence with TechCrunch. The project was solidified after the duo encountered a thought-provoking blog post by Max Leiter, which riffed on Terry Bisson’s classic short story, "They’re Made Out of Meat." The story, which posits that humans are essentially biological machines, served as the conceptual North Star for In the Weights. The creators realized that, in a sense, human history is currently being compressed into floating-point numbers within the latent space of massive neural networks.
The site launched with little fanfare, yet it quickly captured the public imagination. Within weeks of its release, the platform transitioned from a niche developer experiment into a viral curiosity, prompting users across the tech spectrum to check their "strength scores" and compare their digital footprints against celebrities, colleagues, and historical figures.
Supporting Data: How the Weights Are Measured
The methodology behind In the Weights is deceptively simple but technically rigorous. The platform functions as a query-based benchmark. When a user enters a name, the site dispatches a standardized prompt to a variety of models, including industry leaders like Gemini, Claude, and multiple iterations of GPT, as well as smaller, open-weights models like Llama.
The prompt is typically phrased as: "Who is [Name]? Give up to 10 results, each with a short description and confidence."
Once the responses are gathered, the system clusters similar descriptions together and assigns a numerical "strength score." This score is not a metric of popularity in the traditional sense, but rather a reflection of the model’s internal certainty. A high score suggests that the individual’s biography is a well-integrated component of the model’s training data, whereas a low score implies that the AI has either "forgotten" the person or has only a hazy, high-variance understanding of who they are.
The results are often startling. As of late June 2026, the leaderboard reflects a curious blend of pop-culture icons and historical figures. Macaulay Culkin, for instance, has held the top spot with a strength score of 988, standing nearly shoulder-to-shoulder with the late Luciano Pavarotti. Meanwhile, for the average tech professional, scores fluctuate wildly. A score of 641 might land someone in the top 6% of indexed individuals, yet a single update to a model’s architecture or a shift in the training corpus can cause these numbers to swing, revealing the volatile nature of "AI memory."
The "Hallucination" Factor and Model Bias
One of the most valuable aspects of the platform is its transparency regarding model fallibility. By showing which models returned which specific answers, In the Weights exposes the inherent "hallucinations" and biases embedded in LLMs.
For instance, when queried about certain individuals, models sometimes struggle to differentiate between people with similar names or, conversely, may invent details where their training data is thin. During testing, one version of GPT-5.4 Mini described a tech blogger as an "ambiguous name form that could refer to multiple people with the initials A.H.A." This highlights a critical truth about modern AI: these models are not infallible encyclopedias. They are probabilistic engines, and when their "weights" are not sufficiently tuned to a specific entity, they may default to generic or incorrect outputs rather than admitting ignorance.
Dimson has indicated that he plans to expand the platform to analyze these discrepancies further. Future updates aim to explore why specific models exhibit biases toward certain demographics or professions, and to identify "hidden gems"—individuals who have significant contributions to their fields but lack the Wikipedia presence that typically anchors an AI’s knowledge base.
Official Responses and External Critique
The reception to In the Weights has been, in Dimson’s words, "insane." While the creators initially viewed it as a "mild curiosity," the public reaction suggests that the project tapped into a collective anxiety regarding digital legacy.
However, the project has not been without its detractors. Critics, such as AI researcher Anthony Moser, have questioned the utility of the exercise. Moser famously scoffed at the platform on social media, suggesting that it is "literally the same as asking 13 chatbots to tell you about yourself." From this perspective, the "strength score" is merely a proxy for how much training data an AI consumed from the open web—a circular loop where the internet defines the AI, and the AI, in turn, defines the internet.
Despite these critiques, the project remains a fascinating case study in human-computer interaction. Its design—which favors a Nintendo-inspired, retro aesthetic—serves as a soft entry point for what is, at its core, a complex inquiry into data science and information theory.
The Implications: Is This Digital Immortality?
The overarching question posed by In the Weights is whether being "remembered" by a chatbot constitutes a new form of digital immortality. If a person’s life story, career milestones, and personal details are woven into the fabric of a model that will serve as the primary interface for human knowledge, have they achieved a level of permanence that the static, decentralized web could never provide?
The answer is complex. Being "in the weights" is not synonymous with being "known" in the human sense. It is a mathematical representation of a persona. As we move further into a future where AI serves as the primary gateway to information, those who are not present in these models may find themselves effectively erased from the public record. Conversely, those who are present may find their identity permanently altered by the model’s own biases and internal logic.
Looking ahead, the platform serves as a mirror for our own digital identities. It challenges us to consider not just how we are perceived by our peers, but how we are perceived by the machines that are increasingly shaping our reality. As Dimson continues to refine the tool, In the Weights will likely remain a key site of interest for anyone wondering just how much of themselves has been absorbed into the architecture of the future.
Whether it is a definitive measure of relevance or just a clever way to pass the time, the project successfully highlights the most important shift in the digital age: we are no longer just building websites; we are building minds, and we are all, in some small way, trying to see if we have left a mark on the silicon.
