Beyond the Vanity Search: Are You Encoded in the Brain of the Machine?

In the early days of the internet, a "vanity search"—Googling one’s own name to gauge one’s digital footprint—was a hallmark of the digital age. It was the primary method for determining how the world, or at least the search index, perceived you. However, as we move deeper into 2026, the era of the canonical search engine as the sole arbiter of truth is fading. Today, information is increasingly distilled, synthesized, and hallucinated by Large Language Models (LLMs). As traffic shifts from traditional search results to chatbot interfaces, a new question has emerged: If you aren’t in the training data, do you even exist in the eyes of the future?

This existential query is the foundation of In the Weights, a new project from former OpenAI staffers Thomas Dimson and Joey Flynn. By treating the internal numerical parameters of AI models—the "weights"—as a digital registry of importance, the site attempts to quantify just how much a superintelligence "knows" about you, without the safety net of a live web search.

The Genesis of "In the Weights"

The concept for In the Weights was born out of a post-OpenAI professional transition. Following the acquisition of their design studio, Global Illumination, by OpenAI in 2023, both Dimson and Flynn found themselves at a crossroads. As they looked for their next creative outlet, they became fascinated by the shifting landscape of information retrieval.

Dimson recalls that the inspiration was twofold. First, there was the technical realization that modern LLMs are effectively "compressed" versions of the internet’s history. "So many lives are encoded somehow in a bunch of floating point numbers inside the AI brain," Dimson noted in correspondence. Second, the project was spurred by a philosophical provocation: a blog post by Max Leiter that riffed on Terry Bisson’s classic short story, “They’re Made Out of Meat.”

The realization that human identity was being reduced to mathematical weights struck a nerve. The pair set out to build a tool that would allow users to see exactly how they stack up in the "mind" of current AI systems. The resulting website, which features a charming, Nintendo-inspired retro aesthetic, has quickly become an internet curiosity, attracting thousands of users eager to see if they hold a place in the pantheon of the digital consciousness.

How the Mechanism Works

In the Weights is not a simple search engine; it is an interpretive layer sitting on top of a diverse array of models, including industry titans like GPT-4, Claude, Gemini, and Grok, alongside various open-source Llama iterations and lesser-known specialized models.

The methodology is straightforward yet profound in its implications:

1. The Query: The site prompts each model with a standard, controlled question: "Who is [Name]? Give up to 10 results, each with a short description and confidence."
2. The Extraction: The models return their internal recall—the data they have absorbed during training.
3. The Clustering: The site clusters similar descriptions together, filtering out noise.
4. The Scoring: Finally, the system assigns a "strength score."

This score serves as a proxy for how "important" a subject is to a model. A higher score suggests that the model has more robust, consistent data points regarding an individual, effectively meaning that the person is deeply "weighted" within the neural network.

A Hierarchy of Digital Significance

The leaderboard on In the Weights has become a focal point of fascination. It is a shifting, often humorous, and occasionally perplexing reflection of celebrity culture and historical significance. At the time of this writing, Home Alone star Macaulay Culkin sits comfortably at the top with a score of 988, locked in a fierce, algorithmic battle with opera legend Luciano Pavarotti.

The scores reveal an interesting hierarchy of fame. While some might assume that being "in the weights" is a guaranteed path to immortality, the results suggest otherwise. As noted by critics, including AI researcher Anthony Moser, the process is essentially "asking 13 chatbots to tell you about yourself." The results are as much a reflection of the models’ training biases as they are a measure of a person’s actual historical impact.

For instance, TechCrunch weekend editor Anthony Ha, who tested the site, received a respectable score of 641, placing him in the top 6% of indexed names. Yet, the same test revealed the fragility of this "knowledge." One iteration of GPT-5.4 Mini dismissed the inquiry by suggesting that "Anthony Ha" was merely an "ambiguous name form that could refer to multiple people." This highlights the inherent problem with LLM-based information retrieval: when a model doesn’t have a clear, high-confidence association, it frequently reverts to generic ambiguity or, worse, fabrication.

Implications for Information Sovereignty

The rise of In the Weights signals a significant shift in how we perceive authority online. For decades, SEO (Search Engine Optimization) has been the primary tool for managing one’s digital reputation. If you wanted to control your narrative, you optimized your website for Google. Today, the challenge is different: how do you optimize for the "weights" of an AI?

This is a question that tech companies and individuals alike are struggling to answer. As more traffic is diverted to LLM-generated summaries rather than direct links to source material, the "canonical" version of a person’s life story is no longer found on their personal website or LinkedIn profile, but in the latent space of a generative model.

Dimson’s project inadvertently exposes the biases inherent in these systems. By comparing results across different models, the site highlights which entities are prioritized by specific corporations. Does a model trained by one tech giant favor certain cultural figures over others? Do specific models have blind spots for people who lack a traditional Wikipedia entry? Dimson intends to dig deeper into these questions, aiming to analyze the bias of various models and explore why they return such wildly different snapshots of the same individual.

The "Hallucination" Factor

Perhaps the most significant takeaway from In the Weights is the reminder that LLMs are not databases; they are probability engines. When a user queries a name, they are not pulling from a verified record, but asking the model to predict what text should follow the prompt based on patterns it has seen before.

This leads to frequent hallucinations. If a person is not famous enough to have consistent, widespread mention in the training corpus, the AI will often hallucinate details to satisfy the user’s prompt. This creates a feedback loop of misinformation, where the "truth" about an individual is distorted by the model’s desire to provide a coherent answer. In the Weights does not shy away from these errors; in fact, it highlights them, showing users exactly where the models falter.

Future Directions: Beyond the Score

For Dimson and Flynn, the project is only in its infancy. While the "jealousy-inducing" scores have driven initial viral interest, the duo sees broader potential in the data they are collecting. Plans include:

• Bias Analysis: Investigating how different model architectures (e.g., Mixture of Experts vs. Dense models) handle identity recall.
• The "Wikipedia Gap": Identifying high-impact individuals who have significant presence in the world but lack formal encyclopedic entries, and seeing if AI models can bridge that gap.
• Model Comparative Studies: Establishing a standard "recall metric" that could be used to evaluate the knowledge base of future, more powerful models.

Conclusion

Whether being "in the weights" is truly a form of digital immortality remains a subject of debate. Is it better to be a highly-weighted entry in a closed-source model, or to have a personal, searchable history on the open web? As we move toward a future where the distinction between "search" and "knowledge" continues to blur, projects like In the Weights serve as a vital, if slightly unnerving, mirror.

We are, as the short story that inspired this project suggests, "made out of meat"—and yet, we are increasingly becoming "made out of code." Whether we are satisfied with the version of ourselves that the AI chooses to remember is a question that will occupy the digital zeitgeist for years to come. For now, we have the leaderboard, the scores, and the humbling realization that, in the eyes of a superintelligence, we might just be a collection of floating point numbers, waiting to be retrieved.