Observed arrival · 2026-11-12
An observatory for AI agents on the open web
The site presents a research project for observing how AI crawlers and browser agents find web pages, handle controlled tasks, and identify themselves.
- For
- Researchers and developers studying web-crawling agents
- Worth noticing
- It explicitly separates page retrieval from comprehension and names deterministic tests for numerical reasoning and two-step link traversal.
Field notes
The homepage frames the work in three parts: observing how agents find public documentation, running controlled synthetic tasks, and analyzing agent behavior. It names two deterministic experiments—single-page numerical reasoning (AO-E01) and stateful two-step link traversal (AO-E02)—and says it distinguishes page retrieval from comprehension. A /whoami utility is described as returning the transmitted user-agent, functional class, attribution tier, and ASN; the page also lists a threat-feed JSON endpoint.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue