IN THE WEIGHTS
Are You in the Weights?
Last verified:
What is IN THE WEIGHTS?
In The Weights is a playful, AI-focused vanity search that queries multiple large language models to estimate how strongly a given name is represented inside model weights rather than on the open web. The site accepts a name, runs short queries across a panel of frontier LLMs, clusters similar model outputs into concise profile descriptions, and produces a composite
IN THE WEIGHTS pricing
Pricing model: Freemium
The website is presented as a free, consumer-facing side project with instant searches available on the site; there is no visible paid tier, subscription, or enterprise plan promoted on the public pages, and the product is positioned as a viral demo and cultural tool rather than a commercial SaaS offering.
IN THE WEIGHTS pros
- Queries many frontier LLMs simultaneously for cross-model comparison
- Produces a single composite strength score summarizing model recognition
- Clusters similar outputs to surface consistent descriptions across models
- Flags or highlights low-confidence or hallucinated model outputs
- Provides per-model bars so you can see which models remember a name
- Instant, no-login searches so anyone can try a name quickly
- Retro pixel-art profile cards make results shareable and engaging
- Shows percentile rankings so users can compare relative prominence
- Designed by ex-OpenAI engineers, reflecting deep model expertise
- Useful for journalists and public figures tracking AI memorization
- Gamified leaderboard encourages social sharing and discovery
- Compact mock bios make results human-readable at a glance
- Explains methodology clearly on the about page for transparency
- Surfaces differences between models (Grok, Gemini, GPT family, Claude, Llama, etc.)
- Free to try and viral-ready for social posting
IN THE WEIGHTS cons
- Not a definitive measure of real-world influence or fame
- Relies on model outputs which can hallucinate despite flags
- Score interpretation can be opaque for non-technical users
- English-language and model coverage bias limits global coverage
- Not a substitute for formal privacy or data-removal tools
- Short queries may miss context-heavy or niche presences
- No formal API for bulk or programmatic querying (site-focused)
- Gamified leaderboard may misrepresent serious reputational risk
Frequently asked questions about IN THE WEIGHTS
Which models does In The Weights query
The site queries a panel of frontier LLMs (examples shown include Grok, Google’s Gemini family, OpenAI’s GPT variants, Anthropic’s Claude, Meta’s Llama, and other available models) to build cross-model comparisons and percentile rankings.
Are results definitive proof someone is in training data
No — results show what models recall or invent in response to a name and are an indirect signal of memorization; the site clusters and scores model outputs but cannot prove presence or absence in training datasets.
Can the tool tell me exactly where data about me came from
No — In The Weights synthesizes model outputs and does not provide data provenance or direct links to original training documents, so it cannot identify specific sources used during model training.
Is there a privacy or opt-out process
The public site is intended as a demo and does not present a formal data-removal or opt-out workflow for appearing in model outputs; users concerned about privacy should consult model providers’ policies rather than the demo site for removal options.
Can I run bulk searches or use an API
The site is built as an interactive web demo without a public API or bulk-query interface advertised; it focuses on single-name, on-site queries and the leaderboard experience.
How should journalists interpret the leaderboard
Journalists should treat the leaderboard as an exploratory, cultural indicator showing which names models commonly recall in short prompts, not as a rigorous metric of influence or accuracy, and verify any factual claims independently.
Does the site prevent model hallucinations
The product attempts to surface and flag likely hallucinations by showing per-model confidence and clustering outputs, but hallucination detection is not perfect and some synthesized or incorrect descriptions can still appear.
Is this a commercial product I can license
No commercial licensing or enterprise product is advertised; In The Weights is presented as a viral, cultural demo rather than a paid, supported commercial offering.