How does In the Weights work?

In the Weights takes a name, queries several large language models in parallel without web search, clusters their responses to identify which person each model is describing, and outputs a strength score from 0 to 996. Higher scores mean a model can reliably recall the person from training data alone; the top score of 996 is reserved for figures like Mozart, Shakespeare, and Taylor Swift.

Who built In the Weights?

In the Weights was built by Joey Flynn and Thomas Dimson, two former OpenAI employees. They posted it as a Show HN on June 18, 2026, where it reached 430 points and 236 comments. The site frames itself as a way to make 'training data memorization' tangible for anyone who wants to check whether they appear in modern LLM weights.

Is In the Weights accurate?

In the Weights is a probe, not a verifier — the creators acknowledge models fabricate biographical details, typos hurt scores, and common names confuse disambiguation. Smaller models hallucinate more than frontier models. Users in the HN thread report it scores academics with papers and open-source contributors reasonably well but invents biographies for unfamiliar names, especially athletes.

What can I learn from In the Weights?

In the Weights shows which LLMs encode you into their weights and at what strength, exposing how rare names tip into confabulation and how memorization scales with model size. Researchers can use it to study training-data inclusion, writers can check whether models will recognize a quote attribution, and developers can see what an LLM can answer about a person before any retrieval.

In the Weights · 2026-06-18 · major

In the Weights — ex-OpenAI tool scores whether AI models remember your name

Joey Flynn and Thomas Dimson, both ex-OpenAI, launched a free site that types a name into multiple LLMs in parallel and returns a 0–996 'strength score' for how well models recall the person from training data alone.

'In the Weights' homepage banner — query LLMs to see if they recall a name

In the Weights queries multiple LLMs in parallel and scores how strongly each model remembers a person you name.

Quick facts

Built by	Joey Flynn and Thomas Dimson (ex-OpenAI)
What it does	Probes whether a person appears in LLM training data
Score range	0–996 (996 = Mozart / Shakespeare / Taylor Swift tier)
Method	Parallel queries across frontier and small models, clustered
Price	Free
HN	430 points, 236 comments (Show HN)

What is it?

In the Weights is a free web tool built by Joey Flynn and Thomas Dimson, both former OpenAI engineers. You type a person's name and the site reports whether — and how strongly — major language models can recall that person from training data alone, without any web search.

How does it work?

In the Weights sends the name to several frontier and smaller models in parallel, clusters the responses so it can tell which person each model is describing, and combines the results into a single 'strength score' from 0 to 996. The maximum is reserved for names like Mozart, Shakespeare, and Taylor Swift. The creators note that smaller models hallucinate more and that common names confuse disambiguation.

Why does it matter?

In the Weights makes 'training data memorization' tangible. Researchers can probe what LLMs encode about real people, writers can check what a model will recall before relying on it for biographies, and developers can see how much grounding their app needs before a retrieval layer is required. The Show HN hit 430 points in a day, signaling sustained interest in the question of who is — and isn't — in the weights.

Who is it for?

AI researchers, writers, and developers curious about LLM memorization

Frequently asked questions

How does In the Weights work?: In the Weights takes a name, queries several large language models in parallel without web search, clusters their responses to identify which person each model is describing, and outputs a strength score from 0 to 996. Higher scores mean a model can reliably recall the person from training data alone; the top score of 996 is reserved for figures like Mozart, Shakespeare, and Taylor Swift.
Who built In the Weights?: In the Weights was built by Joey Flynn and Thomas Dimson, two former OpenAI employees. They posted it as a Show HN on June 18, 2026, where it reached 430 points and 236 comments. The site frames itself as a way to make 'training data memorization' tangible for anyone who wants to check whether they appear in modern LLM weights.
Is In the Weights accurate?: In the Weights is a probe, not a verifier — the creators acknowledge models fabricate biographical details, typos hurt scores, and common names confuse disambiguation. Smaller models hallucinate more than frontier models. Users in the HN thread report it scores academics with papers and open-source contributors reasonably well but invents biographies for unfamiliar names, especially athletes.
What can I learn from In the Weights?: In the Weights shows which LLMs encode you into their weights and at what strength, exposing how rare names tip into confabulation and how memorization scales with model size. Researchers can use it to study training-data inclusion, writers can check whether models will recognize a quote attribution, and developers can see what an LLM can answer about a person before any retrieval.

Try it

https://intheweights.com/