I-XRAY: The Demo That Turned Wearable Recording Into Real-Time Identification

In October 2024, two Harvard students, **AnhPhu Nguyen** and **Caine Ardayfio**, demonstrated **I-XRAY** — a pipeline that identified strangers in real time using commercially available smart glasses and existing services. It was built and publicised as a warning, and it changed the terms of the smart-glasses debate. ## What it did The pipeline chained together components that already existed: 1. Ray-Ban Meta glasses **livestreamed** to Instagram. 2. A program monitored the stream and detected faces. 3. Detected faces were run through **PimEyes**, a reverse face-search engine that finds other online images of the same person. 4. Results were fed to a **large language model**, which scraped and collated the linked pages and public databases. The output, within seconds, was a stranger's name and details drawn from public sources — reportedly including home address, phone number and relatives' names. ## Why it mattered Nothing in it was novel. That was the point. No new capability was invented; existing consumer hardware and commercial services were connected, and the result was live identification of anyone the wearer looked at. This collapsed the distinction the debate had rested on. Recording someone is one thing; **identifying** them is another, and the second had always been protected by the practical difficulty of matching a face to a name. That friction disappeared. See Practical Obscurity: The Privacy You Actually Have in Public. ## What they did with it They did not release the tool — not to the public, to glasses owners, or to developers — and by their account it was not maintained beyond late 2024. They published it as a public-service demonstration that language models had reached the point of automating this kind of scraping at scale, and accompanied it with **instructions for removing yourself from the databases** the pipeline relied on. ## Consequences The demonstration is regularly cited in the policy argument that followed: the ACLU and dozens of other organisations called for facial recognition to be abandoned in these products, courts and venues introduced bans on camera glasses, and biometric-consent statutes such as Illinois' BIPA gained renewed attention. The lasting significance is that it converted a hypothetical into a demonstrated capability built by two students in their spare time — which is a much harder thing to argue against than a hypothetical. See Why Camera Glasses Draw More Backlash Than Phones.