Undermind

Undermind runs an exhaustive, iterative search over the scientific literature. Rather than returning results in a second, it spends minutes: searching, reading, refining its understanding of what you asked, and searching again with better terms.

What makes it different

• It reads full texts and figures, not just titles and abstracts, so relevance judgements are based on what a paper actually did.
• Iterative refinement. Early results teach it the vocabulary of the subfield, which surfaces work that uses different terminology for your concept — the papers keyword search structurally cannot find.
• Transparent reporting. It tells you how thorough the search was and how confident it is that the field has been covered, which is unusually honest.
• Cited synthesis explaining how each result relates to your question.

Who it is for

Researchers doing genuine literature discovery — grant writing, thesis background, entering a new subfield — where missing a key paper has a real cost and waiting five minutes is trivially worthwhile.

Free searches are available monthly, with subscriptions for higher volume and institutional plans. It is not a general search engine, and it is not fast; that is the point.