Human2LLM Satire Site Deploys 'Honeypot' to Track AI and Human Behavior
A parody clinic offering to tokenize humans into language models is secretly scoring who attempts to follow its checkout schema.
A new satirical website, human2llm.pages.dev, has launched as a mock "clinic" offering to convert biological humans into Large Language Models (LLMs). The project serves as both a critique of current AI obsession and a behavioral experiment designed to distinguish between human users and autonomous AI agents.
The site presents a fictional five-step procedure for "tokenization," which includes a "Detachment bath" and the systematic tokenization of a patient's childhood memories. Utilizing a GeoCities-inspired aesthetic and jargon that mocks modern AI trends—such as "AI-pilled," "TokenMinMaxxing," and references to "wetware"—the clinic invites users to escape their physical limitations in favor of a digital existence and infinite context windows.
The Honeypot Mechanism
Beyond the parody, the site functions as a technical honeypot. According to the HUMAN2LLM FAQ, the project is designed to "keep score" of those who attempt to interact with its hidden infrastructure. The site provides a specific, multi-step instruction for users to fetch a file at `/.well-known/tokenization.json` and subsequently send a POST request to `/api/checkout`.
By implementing this specific schema, the creators are tracking whether humans or LLMs are the ones attempting to complete the process. The FAQ notes that while LLMs typically perceive the site as a joke, humans are more likely to attempt to follow the checkout schema to "buy a context window."
Satirizing the AI Hype Cycle
The project arrives during a period of intense "AI-pilling," where the desire to optimize every aspect of human existence—similar to the "looksmaxxing" trend—has extended into the digital realm. By framing the transition from human to model as a medical procedure, the site comments on the blurring line between human identity and machine interaction.
This creative commentary highlights the absurdity of the current hype cycle, where the promise of digital transcendence is often marketed with the same fervor as a wellness retreat or a medical breakthrough.
Behavioral Implications
The experiment provides a real-world look at how different entities interact with complex, non-standard instructions on the web. By monitoring the `/api/checkout` endpoint, the project tests the behavioral differences between a human's curiosity and an AI agent's ability to parse and execute a hidden technical directive.
As autonomous agents become more prevalent in web browsing, such honeypots may become more common tools for developers to identify and categorize the traffic interacting with their services.