Lina Hope
The ritual of "Googling oneself" has long been a staple of modern digital life—a quick, sometimes ego-boosting, sometimes humbling check to see how the world perceives us through the lens of search engine optimization. But in 2026, the cultural landscape has shifted. As generative AI models move from novelty chatbots to the primary gateways for information retrieval, the question is no longer "What does Google say about me?" but rather, "Am I encoded in the machine?"
Enter In the Weights, a provocative new project launched by former OpenAI alumni Thomas Dimson and Joey Flynn. The platform attempts to quantify the intangible: how deeply a person’s identity has been woven into the neural pathways of modern Large Language Models (LLMs). By bypassing traditional web search and forcing models to rely solely on their internal training data, the site offers a startling, if not entirely scientific, look at who, or what, the "AI brain" considers significant.
The Genesis of "In the Weights"
The concept for the site was born from a post-OpenAI period of reflection for Dimson and Flynn. The duo, who joined OpenAI following the acquisition of their design studio, Global Illumination, found themselves contemplating the future of identity in an era of superintelligence.
"Google vanity searches are the wrong objective in 2026 as more traffic moves to LLMs," Dimson noted in a recent correspondence. He and Flynn were struck by the realization that our digital legacies are no longer just links on a page, but are instead "encoded somehow in a bunch of floating point numbers inside the AI brain."
The project gained its philosophical momentum from a satirical blog post that riffed on Terry Bisson’s classic short story, They’re Made Out of Meat. The story’s existential dread—that human consciousness is merely physical matter—is mirrored in the project’s name. "Being in the weights," the site declares, "means your existence was deemed important in the process of creating superhuman artificial intelligence."
How It Works: Probing the Neural Fabric
To determine a subject’s "weight," the platform queries a diverse array of models—ranging from industry titans like GPT-4, Claude, and Gemini to the open-source Llama and various niche, smaller models. The prompt is deceptively simple: "Who is [Name]? Give up to 10 results, each with a short description and confidence."
The system then aggregates these disparate responses, clustering similar descriptions to assign a "strength score." This score is a proxy for how ingrained a person is in the model’s training set. A high score suggests that the model has seen enough data during its pre-training phase to form a coherent, confident, and accurate profile of the individual without needing to consult the live internet.
The results are presented with a nostalgic, Nintendo-inspired aesthetic, adding a layer of playfulness to a fundamentally heavy subject. However, beneath the retro interface lies a complex data-gathering engine that exposes the varying levels of "knowledge" different models possess.
Chronology of a Digital Phenomenon
- Pre-2026: The decline of traditional search engine utility begins as users increasingly rely on conversational AI for direct answers rather than browsing blue links.
- May 2026: Dimson and Flynn finalize the architecture for In the Weights, deciding to focus on "zero-shot" recall—testing the models without allowing them to access web-search tools.
- Early June 2026: The site goes live, initially as an internal experiment among tech circles, before rapidly spreading through social media.
- Mid-June 2026: The leaderboard gains traction. Stars like Macaulay Culkin and cultural icons like Luciano Pavarotti emerge as high-scorers, creating a competitive atmosphere.
- Late June 2026: The project begins to face scrutiny from the AI research community, sparking debates over whether these scores actually reflect "importance" or merely the biases of training data.
Supporting Data and the "Hallucination" Factor
The platform provides more than just a score; it offers a transparent view of the raw responses generated by the models. This transparency has become one of the project’s most valuable features, as it inadvertently exposes the persistent issue of AI hallucinations.
For instance, when queried about TechCrunch editor Anthony Ha, the GPT-5.4 Mini model returned a curious description, identifying him as an "ambiguous name form that could refer to multiple people." While a human would recognize the inaccuracy, the model’s output highlights the fragility of internal knowledge.
The leaderboard itself is a fluid entity, constantly shifting as more queries are logged and as different models are updated. As of this writing, celebrity status and historical significance correlate strongly with high scores, though the "strength" of an entry often depends on the breadth of the training data. A person might be a household name, but if they lack a significant digital footprint in the corpus used to train a specific model, their weight remains low.
The Critical Reception
The reaction to In the Weights has been polarizing. To some, it is a fascinating, almost existential gauge of modern relevance. To others, it is a redundant exercise in vanity. AI critic Anthony Moser dismissed the project on social media, noting that the score is "literally the same as asking 13 chatbots to tell you about yourself."
Despite the skepticism, the public engagement has been "insane," according to Dimson. He suggests that the site has struck a deep-seated human desire to feel "remembered" by the systems that are increasingly defining the modern world. There is an undeniable competitive thrill to the ranking system—a digital version of "keeping up with the Joneses" for the AI era.
Implications for the Future of Identity
The project raises profound questions about the nature of fame and information in the age of LLMs. If the "weights" of a model represent the sum total of its "knowledge," then inclusion in those weights is, in a very real sense, a form of digital immortality.
The Bias of the Model
Dimson has indicated that future iterations of the project will delve deeper into the systemic biases within these models. By analyzing which models favor specific demographics or categories of people, In the Weights could serve as an essential audit tool for AI ethics researchers. Are models biased toward Western celebrities? Do they struggle to represent figures from the Global South? These questions are at the heart of the next phase of the project.
The "Wikipedia" Benchmark
One of the most intriguing avenues for the team is investigating individuals who, by all accounts, "should have a Wikipedia article but don’t." This highlights the gap between public perception and the data that models actually ingest. It suggests that our definition of "importance" is being redefined not by historians or journalists, but by the algorithms that scrape and synthesize our collective digital output.
Beyond Vanity
While the project is currently framed as a playful "vanity" tool, the implications are serious. As AI models become the primary engines for research, education, and even legal discovery, being "in the weights" will become a prerequisite for relevance. Businesses, activists, and public figures will likely start to optimize for "weight" rather than "rank"—a shift that could fundamentally alter how we manage our digital identities.
Conclusion: The Mirror of the Machine
In the Weights is more than just a quirky website; it is a reflection of a society grappling with the loss of its traditional information gatekeepers. In the past, being "found" meant being indexed by a search engine. Today, being "known" means being computed.
As Thomas Dimson and Joey Flynn continue to iterate on their creation, the site will likely serve as a barometer for how these models evolve. Whether or not you believe that being stored in a model’s parameters is a path to immortality, the project forces us to acknowledge a new reality: we are no longer just living our lives; we are being processed. For better or worse, we are all, in some measure, becoming "meat" in the machine.
As the leaderboard continues to shift, one thing is clear: the quest to see how we stack up against the competition isn’t going anywhere. It has simply migrated to a new, more mysterious, and infinitely more complex digital frontier. The question remains: when you look into the "AI mirror," do you like what you see, or is the reflection just a hallucination?
Siti Muinah
Layla Zulfa
