Thomas Dimson and Joey Flynn launched In the Weights, a site that scores how well large language models recall a person by name, turning the old habit of Googling yourself into a chatbot leaderboard. The site queries 13 models — including Grok, Gemini, multiple versions of GPT, Claude, and Llama, plus several lesser-known systems — and asks each one to identify a given name with up to 10 candidate descriptions and confidence ratings. It then clusters similar answers and assigns a single strength score.
Macaulay Culkin currently sits at the top with a score of 988, running neck-and-neck with opera singer Luciano Pavarotti. The leaderboard moves in real time as new names are submitted and scored, and the public results page shows which model returned which description, surfacing disagreements and hallucinations across systems.
“Being in the weights means your existence was deemed important in the process of creating superhuman artificial intelligence”— In the Weights, Site description
The pitch is part diagnostic, part vanity mirror. A score of 641 places a name in the top 6% of those queried — roughly the level the site assigned to TechCrunch's Anthony Ha, who wrote up his own result. The premise is that web search no longer captures how people actually look each other up, with a growing share of identity lookups now happening inside chatbots rather than on Google.
Key facts
- 01In the Weights queries 13 models — including Grok, Gemini, multiple GPT versions, Claude, and Llama — to score name recall.
- 02Macaulay Culkin currently tops the leaderboard with a strength score of 988, neck-and-neck with Luciano Pavarotti.
- 03The site clusters similar model responses and assigns a strength score; a 641 placed one TechCrunch writer in the top 6% of names.
- 04Creators Thomas Dimson and Joey Flynn came to OpenAI via the 2023 acquisition of their design startup Global Illumination.
- 05GPT-5.4 Mini surfaced as a hallucination flag, calling 'Anthony Ha' an ambiguous name form for multiple people with initials A.H.A.
Dimson and Flynn came to OpenAI through the acquisition of their design startup Global Illumination, and built In the Weights after leaving. Dimson told TechCrunch by email that the project was an attempt to get the creative juices flowing again, and that the framing crystallized around a tongue-in-cheek blog post riffing on AI weights and Terry Bisson's short story "They're Made Out of Meat."
The underlying argument is sharper than the toy suggests. Dimson said he had been thinking about how vanity searches on Google look like the wrong objective in 2026 as more traffic moves to LLMs, and about the fact that so many lives are now encoded somehow in floating point numbers inside model weights. Whether a person shows up in those weights — and how consistently — is increasingly a proxy for online presence in a world where users ask a chatbot instead of opening a search results page.
“Google vanity searches are the wrong objective in 2026 as more traffic moves to LLMs”— Thomas Dimson, Co-creator of In the Weights
The site also exposes the messy reality of model recall. GPT-5.4 Mini described Anthony Ha as an ambiguous name form that could refer to multiple people with the initials A.H.A. — a hallucination dressed up as analysis. Different models in the same family return different results for the same name, and the public scoreboard makes those splits legible in a way private chatbot conversations do not.
Dimson said the response surprised him. He and Flynn had expected a mild curiosity and instead found the site striking a nerve around whether people live forever in superintelligence. The comparison-and-leaderboard mechanic, he conceded, did not hurt.
“Reception has been insane so far, we thought this would be a mild curiosity but it seems like it has struck a nerve of wanting to see if you live forever in the super intelligence”— Thomas Dimson, Co-creator of In the Weights
Not everyone is impressed. AI critic Anthony Moser scoffed that the exercise is literally the same as asking 13 chatbots to tell you about yourself — which, mechanically, is fair. The aggregated strength score is a presentation layer on top of repeated prompts, not a novel evaluation method. The site also leans on a retro, Nintendo-inspired design that makes the output feel more authoritative than the underlying queries warrant.
Dimson said he plans to dig further into why models in the same series return different results for the same name, which models bias toward which kinds of people, and which people should have a Wikipedia article but don't. That last thread points at the real research question buried under the vanity hook: training-data coverage is uneven, and the unevenness has consequences for who shows up when a user asks a chatbot about a person, a company, or a place.
The site arrives at a moment when ChatGPT's share of the chatbot market has slipped below 50% for the first time, and when identity lookups are fragmenting across half a dozen assistants with different training cutoffs and different retrieval stacks. In the Weights is a small piece of software, but it surfaces a structural question every model vendor will have to answer: name recall is becoming a reputation primitive, and the model weights are the new index. Expect SEO-style optimization for LLM presence to follow, with all the gaming, manipulation, and arbitrage that implies.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
