EVIDENCE / READ THE TESTS
A useful premise should survive inspection. Start with the question you care about, then read the method, result and limits together.
HAUSE publishes evaluations of form selection, question routing and what models learn from its website. These are scoped tests by the author, not proof of broad adoption. VINDEX3 and CHRISHAYUK demonstrate use in two publications by the same author; independent production reuse has not yet been established.
CHOOSING-1
Three selectors face the same unfamiliar content: names and descriptions, the full selection grammar, and the deployed deterministic resolver. Read the scores and the failure analysis together.
READ THE EVALUATION ↗ROUTING-1
The same vocabulary is used to request different kinds of answers. This tests routing before retrieval or form selection.
READ THE EVALUATION ↗ROUTING-2
A follow-up tests moving selection from keyword scaffolding to the choosing records. It separates unfamiliar questions from phrasings already encountered.
READ THE EVALUATION ↗READING-1
Fresh model contexts read the public corpus and answer eight questions with sources. The record examines the understanding the site creates, including its omissions.
READ THE EVALUATION ↗READING-2
Two sentences are added to the homepage and the reading exercise is repeated. The record follows whether the boundary travels, and which sources readers use.
READ THE EVALUATION ↗ADOPTION & POSITIONING
The case studies show how real pages shaped the library. A production implementation by another author is the next adoption proof point. It is an open task, not a case study this site can claim.