Thomas Dimson and Joey Flynn built In the Weights to show you what AI models remember about you without search. That is a sharper reputational test than a vanity Google search, because the answer now comes from the model before it ever sends anyone to a link.
If you still Google your own name to see how the world sees you, you're checking the old front door. As TechCrunch reported, Thomas Dimson and Joey Flynn, who previously worked on Instagram's engineering and design teams and later joined OpenAI through its 2023 acquisition of Global Illumination, have launched intheweights.com, a free site that tests how strongly large language models can recall you from their own weights.
The name is the point. In the Weights asks models including GPT-5.5, Claude Opus 4.8, Gemini, Grok, Llama, and smaller systems what they know about a person, then turns that into a strength score. It isn't measuring how well your website ranks. It isn't asking whether your LinkedIn page has the right keywords. It is asking whether the model already has enough stored signal to identify you, describe you, and place you in the right context without going out to search.
Also read: Microsoft is rewriting the economics of enterprise AI and the bill shock is just getting started • Stanford's 2026 AI Index confirms the enterprise window is closing faster than most founders think • Kevin Warsh's Fed debut leaves startups and AI borrowers navigating a world without rate cut relief
That sounds narrow until you think about how people now use AI. A hiring manager doesn't always start with ten blue links. A customer researching a consultant may ask a chatbot for a short read. A founder, journalist, investor, researcher, or creator can be summarized by a model before anyone clicks through to the original source. If the model gets you wrong, leaves you out, or confuses you with someone else, your search ranking doesn't fix the first impression.
Dimson and Flynn are a sensible pair to build this. Global Illumination was an AI design and product startup, and Reuters reported in August 2023 that OpenAI acquired it to work on core products including ChatGPT. Before that, Dimson helped build Instagram's ranking and discovery systems, the kind of work that trained a generation of product people to care about how people and posts are surfaced. Flynn's design background matters too, because this problem isn't only technical. It is about what a person sees when a model compresses a career into three sentences.
Here's the thing: reputation used to be something you could inspect. You could search your name, scan the first page, update a bio, publish a correction, ask a site to fix a mistake. AI makes that more awkward. The answer is generated, not listed. The model may draw from old writing, press mentions, public profiles, forum posts, code repositories, podcasts, conference pages, and whatever else made it into training or retrieval systems. You don't get a neat index of every influence. You get an output.
In the Weights is useful because it turns that vague anxiety into something you can actually test. If several models know you accurately, that tells you your public record has enough durable signal. If only one model recognizes you, or if the model invents credentials you never had, that is not a cosmetic issue. It tells you the AI layer between you and the reader is already making choices about your identity.
Don't overstate the tool. A score from In the Weights is not a complete reputation audit, and it won't tell you why every model answered the way it did. Models also change. A person who appears clearly in one release can blur in another after a retraining cycle, a retrieval change, or a safety update. But the basic question is real: when AI systems answer on your behalf, do they know who you are?
For startups, that question is no longer vanity. Founders spend years trying to become findable to investors, customers, job candidates, and reporters. They tune websites, write posts, appear on podcasts, and chase backlinks because Google made visibility measurable. AI search breaks that habit. You can have a clean website and still be thin in a model's memory. You can also have years of messy public material and discover that a chatbot is now the one doing the editing.
That is why this little site matters. It gives you a quick look at a new kind of public surface, one you don't fully control but can't afford to ignore. Google told you what pages ranked. In the Weights tells you whether the machines can place you at all.
