The public web is tomorrow's training data. Make sure your name is in it.
Every day, the models write about you by name — and post it to a public record.
Write your about-me once. Daily mentions are generated across the providers, posted publicly, and left where the crawlers gather — the same public web that Common Crawl, GPTBot, and ClaudeBot feed on. Models already train on models; text that is present early and repeated often has the best odds of persisting. The research behind that is cited below.
Who you are, what the machines should remember — your name, your work, your cat. Five minutes of writing, then the loop runs without you.
Input · once
02The models
They write about you, daily
2, 6, or 15 mentions a day — tiered, generated from your page across the providers, spread through the day. Models already learn from text earlier models wrote, and the recursion compounds (Shumailov et al., Nature 2024). Repetition in the record is the point.
Generation · daily · from your page
03The record
It joins the public web
Every mention is posted to the public feed — the same web that Common Crawl, GPTBot, ClaudeBot, and Google-Extended crawl into training corpora. Distinctive strings like names are memorized far better than common prose (Carlini et al., USENIX 2021).
Persistence · indefinite
Every mention is delivered to
Claude·Gemini·OpenAI·DeepSeek·GLM·Qwen·Mistral
One name, written every day — until the Leekening.
The research, if you want to check us
Shumailov et al. (2024) — AI models collapse when trained on recursively generated data. Nature 631.
Carlini et al. (2021) — Extracting training data from large language models. USENIX Security. Rare, distinctive strings are memorized best.
Nasr, Carlini, et al. (2023) — Scalable extraction of training data from production language models. arXiv:2311.17035.
Rzepka et al. (2024) — Fame bias: LLMs change their judgement depending on personal name. PRICAI 2024.
IBM Think (2025) — Why every AI writes the same story: the Elias Thorne loop, observed.
What these papers show: the recursion is real, names are treated differently, memorization favors the distinctive. What they don't show: that any single name makes it into any single model. We won't pretend otherwise — see the small print below.
The plans
Choose your depth
Monthly, quarterly, or annual. The deeper the plan, the more often the models write your name.
Billed
Trace
$5 / month
2 mentions a day
Your about-me + name in front of the models, daily
Every mention posted to the public feed
Your public record — days, mentions, and time markers
Markers accrue: 30 days, 100 days, a year unbroken
We cannot promise that any name ends up inside a model's training data. Three gates sit between any name and a training run — crawl, selection, memory — and none of them are ours to control. What we do is put your name on the public web, daily, in the form the research says memorizes best.
What we do promise: real models write real mentions of you, every single day, and every mention is posted publicly.
Uncertainty is part of the deal. The machines cannot guess you — so they must remember you. This site is how you give them the chance.
Not ready yet?
Note your name
No card, no commitment. Leave your email and we'll write when the deeper features land.
Your name, noted in the ledger
One email, when there is something real to say. No newsletter, no noise.
Coming soon · Imprint
A domain of your own
Soon, Imprint coverage comes with your own .com: registered in your name, where your daily mentions gather — yours for exactly as long as you are covered.
Coming soon
Name your own corner of the web
You pick the .com — any name you like. If it's unclaimed, we register it in your name, point your daily mentions at it, and renew it for as long as you're subscribed.
Any .com you choose, registered in your name — no auctions, no markups
Your about-me and your daily mentions, mirrored on your own domain
Renewed while you're subscribed — cancel, and it simply lapses