About the Role

We are building an evaluation benchmark for frontier AI browsing agents. Your job is to design research problems that a state-of-the-art AI cannot solve, even with full web access and multiple attempts. This is not a subject matter expert role, nor is it a content-writing role. It is investigative research.

You will start from a verifiable fact, work backwards to construct a question that makes that fact extremely hard to locate, and then prove your work with a complete, auditable evidence trail.

What You Will Produce

A natural-language research question with a short, stable, objectively verifiable answer
Clues, each independently checkable, spanning multiple fact types - including dates, people, places, organisations, works, events, records, and quantities - with specific constraints
A validation record showing the obvious searches you ran and the results they returned

Minimum Qualifications

Demonstrated open-web research ability, including locating primary records and navigating government and institutional databases, archives, registries, and PDF documents
Precision with sourcing. You cite exact pages, tables, and sections—not just homepages
Comfort researching unfamiliar subjects from scratch
Native or near-native written English
High tolerance for structured documentation. The evidence trail is the majority of the work
Experience with LLM evaluation, red-teaming, or benchmark construction
Experience in one or more of the following domains:

Reference librarianship, archival research, or special collections

Investigative journalism or professional fact-checking

OSINT, due diligence, KYC, or investigative research

Patent, prior-art, or legal-discovery search

Genealogy and records research

Competitive quizzing or puzzle-hunt construction

Nice to Have

Experience with LLM evaluation, red-teaming, or benchmark construction
Familiarity with JSON and structured data delivery formats

Web Research Specialist