In the Weights Turns AI Memory Into the New Vanity Search

A new website called In the Weights is turning a familiar internet habit into an AI-era experiment: searching for yourself.

For decades, vanity search meant typing your own name into Google and seeing what appeared. The results could show old articles, social profiles, company pages, public records, blog posts, interviews, images, or random mentions across the web. It was a quick way to measure visibility in the search-engine age.

In the Weights asks a different question. Not what the web says about you, but whether AI models seem to remember you without using the web at all.

The project, created by Thomas Dimson and Joey Flynn, is built around the idea that large language models now shape public visibility in a new way. If a model can describe a person from its internal knowledge, without live search or external tools, that person may be “in the weights.” In simple terms, their public presence may have been absorbed deeply enough into the model’s training that the AI can recall them from memory.

The site is playful, but the idea is serious. As AI assistants become a new way people search, summarize, and discover information, being visible on the open web may no longer be the only marker of digital relevance. The next question may be whether AI knows who you are.

Search Is Moving From Links to Memory

Traditional search is built around retrieval.

A search engine takes a query, scans an index of web pages, ranks results, and sends users toward links. Visibility depends on whether a page exists, whether it is indexed, whether it ranks, and whether people click.

AI search changes that pattern. A chatbot or AI assistant may answer directly, summarizing information instead of sending users to a list of links. Sometimes it uses live web search. Other times, it answers from what the model has already learned during training.

That difference is what makes In the Weights interesting.

The site is not measuring normal web ranking. It is trying to measure model recall. If an AI model can identify and describe a person without browsing, that suggests the person’s public footprint may have become part of the model’s internal statistical structure.

That is the “weights” in the name. In machine learning, weights are the numerical parameters that shape how a model produces output. The site turns that technical concept into a cultural signal: are you important enough for the model to remember?

A New Kind of Internet Status

Vanity search has always been partly about ego.

People search themselves to see whether they appear, how they are described, and whether the internet reflects their identity accurately. In the Google era, this was tied to search ranking. In the social era, it was tied to followers, verification, engagement, and mentions.

In the AI era, status may take a stranger form.

If models can describe a founder, journalist, artist, academic, investor, creator, or executive, that becomes a new kind of visibility. It suggests that a person’s work has circulated widely enough to be represented in training data. The model may not show sources. It may not be fully accurate. But the fact that it can produce a coherent description becomes its own signal.

That is why the site works as a vanity tool. People want to know not only whether Google can find them, but whether AI can remember them.

This could become especially appealing in tech, media, academia, entertainment, and startup circles, where personal visibility already matters. A high “In the Weights” score becomes a new form of reputation measurement, even if the metric itself is imperfect.

The Founders Are Testing AI Recall

Thomas Dimson and Joey Flynn built In the Weights around the feeling that AI memory is becoming culturally meaningful.

The site reportedly prompts multiple models to describe a person without using tools such as web search. It then compares the outputs, groups similar descriptions, and produces a score that reflects how strongly the models appear to recall that person.

That method is not the same as proving what was in a training dataset. A model’s output does not reveal its full training history. It may generate correct details, mix up people, hallucinate, or repeat information from patterns in public data. Still, the outputs can show whether a model has enough internal signal to produce a recognizable profile.

That makes the tool part game, part media experiment, and part commentary on AI visibility.

The score should not be treated as a scientific measure of importance. But it captures something many people are starting to feel: AI systems are becoming a new mirror for public identity.

AI Memory Can Be Wrong

The problem with AI vanity search is that models do not always remember accurately.

A model may confuse two people with similar names. It may invent job titles, mix public facts, exaggerate achievements, or attach someone to work they did not do. It may describe a person confidently even when the details are wrong.

That makes AI memory different from search results. A search page usually shows sources that users can inspect. A model answer can feel polished and authoritative while hiding where the information came from, whether it is current, and how much of it is inferred.

For people with public profiles, this creates a new reputational issue. It is not enough to ask whether information about them is online. They may also need to ask whether AI systems summarize them correctly.

If a model gets a person wrong, correcting the record is not simple. You cannot edit the model’s weights the way you can update a website bio. You may be able to improve public information, request corrections in some products, or wait for future model updates, but the process is less direct than fixing a web page.

The Tool Points to a Bigger Discovery Shift

In the Weights is playful, but it points to a bigger change in how people and companies may think about discoverability.

Search engine optimization was built around ranking in Google. Social media visibility was built around algorithms, shares, and audience engagement. AI visibility may depend on a different mix: reliable public information, repeated mentions across credible sources, structured profiles, media coverage, authorship, entity recognition, and model training patterns.

That creates a new challenge for brands, founders, creators, and professionals.

In the future, the question may not only be “Do I rank on Google?” It may be “What do AI assistants say about me?” and “Do they know I exist without having to search?”

This matters because AI assistants are becoming a first stop for information. Users ask ChatGPT, Gemini, Claude, Perplexity, Grok, and other systems to summarize people, companies, products, and trends. If the answer is wrong or incomplete, it can shape perception before the user ever visits a website.

AI visibility may become a new layer of reputation management.

Being Remembered Is Not Always Good

There is also a darker side to being “in the weights.”

Some people may not want AI models to remember them. Public figures, journalists, activists, academics, executives, and creators may already expect some level of public visibility, but ordinary people may not. Even for public people, being remembered by a model can raise questions about consent, privacy, and control.

The phrase sounds flattering, but it also means personal information may have been absorbed into systems that are difficult to inspect or change. A person may not know what data contributed to the model’s memory, whether it is accurate, or how it will be used.

This creates a tension. In some circles, being known by AI may feel like status. In others, it may feel invasive.

That tension is likely to grow as AI assistants become more common. People will want the benefits of accurate representation, but they may also want limits on how much AI systems remember about them.

The New SEO May Be Entity Management

In the Weights also hints at a future where AI visibility becomes part of professional identity management.

For companies and public professionals, the goal may be to make sure AI systems understand who they are, what they do, and how they should be described. That could push more attention toward structured data, clear official bios, consistent public profiles, authoritative references, and high-quality content.

This is not exactly traditional SEO. It is closer to entity management.

Search engines rank pages. AI assistants often summarize entities: people, companies, products, institutions, events, and ideas. If the model’s understanding of an entity is weak or confused, the answer may be vague or wrong.

That means digital presence may need to become cleaner and more consistent. Fragmented bios, outdated pages, inconsistent job titles, and duplicate names could cause more confusion in AI-generated answers.

For anyone whose reputation matters online, AI recall may become something to monitor.

A Cultural Joke With Real Stakes

The reason In the Weights works is that it turns a technical concept into a social one.

Most users do not think about model parameters, training data, or retrieval systems. But they understand the feeling of asking, “Does the internet know me?” The site updates that question for the AI age.

That makes it funny, but not trivial.

People already care about whether they appear on Google, whether they are verified, whether they have a Wikipedia page, whether their LinkedIn profile ranks, or whether their work appears in AI answers. In the Weights adds a new status test: do models recall you from internal memory?

The answer may be amusing for now. But the underlying shift is real. AI systems are becoming new gateways to information, and their internal representations may influence how people, brands, and public figures are discovered.

AI Search Will Change Reputation

The rise of AI-centric vanity search shows how reputation is being reshaped by artificial intelligence.

The web once rewarded being indexed. Social media rewarded being followed. AI may reward being encoded, summarized, and remembered by models that users increasingly trust as information guides.

That does not mean AI memory is objective or fair. It can be biased, outdated, incomplete, or wrong. It may favor people with more public coverage, stronger English-language presence, or more frequent mentions in training data. It may overlook important people who are less visible online or who work in less documented communities.

But imperfect metrics often become influential anyway. Search rankings were imperfect. Follower counts were imperfect. Verification badges were imperfect. AI recall scores may follow the same pattern.

In the Weights is a small website built around a clever idea, but it captures a larger cultural moment. The internet is moving from “Can people find me?” to “Does AI already know me?”

That question may sound vain. Soon, it may become part of how people understand digital visibility itself.