In the Weights: AI’s New Gatekeepers of Digital Identity

The Invisible Algorithm and Curated Significance

The latest digital novelty, ‘In the Weights,’ presents itself as a playful vanity metric: a score indicating how well large language models (LLMs) like Grok, Gemini, and Llama recall your existence without the crutch of web search. Its creators, Thomas Dimson and Joey Flynn, two former OpenAI employees, frame it as a necessary evolution beyond the fading relevance of traditional Google searches. They suggest that as more traffic shifts to conversational AI, a new form of digital recognition is needed, measuring how ‘lives are encoded somehow in a bunch of floating-point numbers inside the AI brain.’

This framing, however, overlooks the profound implications of moving from an indexed, albeit imperfect, web to an encoded, opaque algorithm. A score of 641, placing one in the top 6% of recognized names, or even Macaulay Culkin’s commanding 988, feels satisfying precisely because it’s a black box. The core issue isn’t whether one’s name appears, but *whose* definition of importance the underlying AI models are trained on and replicate. This platform is not just a search tool; it’s an early, gamified glimpse into the AI industry’s emerging, often biased, power to define and curate human significance through algorithmic encoding.

It’s tempting to dismiss ‘In the Weights’ as mere digital narcissism, a charming retro-styled amusement. But this isn’t merely about ego. The site exposes a crucial, under-examined tension in the AI era: the fight for legitimate digital existence in a world increasingly mediated by proprietary datasets and models. The incentive here is subtle: by offering a glimpse into the ‘weights,’ it reinforces the perceived authority and near-mythical ‘intelligence’ of these closed systems, subtly nudging users to accept AI as the new arbiter of influence and memory. Who benefits? The AI developers themselves, by legitimizing their models as universal consciousness.

The Shifting Sands of Digital Memory and Bias

For over two decades, web search engines, despite their flaws, offered a relatively transparent — if not always equitable — mechanism for information retrieval. Your digital footprint was, in theory, discoverable through links, articles, and public records. The move towards LLM-based recall fundamentally alters this dynamic. Instead of parsing the web in real-time, these models retrieve from their fixed training data, a snapshot of the internet at a specific point in time, curated and weighted by unseen algorithms.

The creators themselves acknowledge they plan to investigate ‘which models are biased towards different types of people.’ This is not a future concern; it is the fundamental reality of their product. When GPT-5.4 Mini offers a generic ‘ambiguous name form’ for an individual like TechCrunch’s Anthony Ha, it reveals less about Ha and more about the gaps, biases, and inherent limitations of the model’s training data. It’s a stark reminder that these models are not omniscient; they are reflections, often distorted, of the data they consume.

The most skeptical observation about ‘In the Weights’ isn’t its vanity, but its pretense of objective measurement. As AI critic Anthony Moser bluntly put it, this is ‘literally the same as asking 13 chatbots to tell you about yourself.‘ That sentiment captures the essence of the problem: a quantitative score based on qualitative, often unreliable, AI output provides a false sense of validation. It transforms the messy, complex reality of human identity into a simple integer, driven by a proprietary, inscrutable process.

Whose Consciousness Are We Feeding?

The fascination with ‘living forever in the super intelligence,’ as Dimson notes, is profound. This taps into an ancient human desire for legacy, now recast through the lens of artificial general intelligence (AGI) narratives. Yet, to be ‘in the weights’ is to be remembered not by humanity, but by a machine whose ‘memory’ is static, susceptible to ‘hallucinations,’ and inherently non-human. It’s a digital necropolis where only certain ‘souls’ are digitized, and even then, often with inaccuracies or omissions.

Consider the broader context: the relentless accumulation of data for AI training, often without clear consent or compensation, forms the bedrock of these systems. Every article, every public profile, every tweet becomes a parameter in a vast neural network. Platforms like ‘In the Weights’ capitalize on this data extraction by turning the consequences of that extraction into a consumer-facing metric. It reinforces the idea that an AI’s internal representation of you is a meaningful measure of your existence, rather than a mere algorithmic artifact.

The real question isn’t how high your score is, but rather, what power structures are being solidified by the very act of seeking this score. As our digital identities increasingly depend on the whims and internal logic of AI models, we risk ceding control over our own narratives to systems designed by a privileged few. This is not about being remembered by a benevolent superintelligence; it’s about being cataloged, scored, and thus defined, by Silicon Valley’s latest set of digital gatekeepers, whose ‘weights’ carry more social currency than we often care to admit.