Map research papers with Claude and a searchable corpus
At a glance
A researcher used Claude and a public research corpus to organize papers for a string theory project. This workflow is listed under Research, Academic Papers, Learning.
- Agents / tools
- Claude Code, ExoPriors
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Start with a research question and Claude Code. The Show HN service offered a prompt containing access to its public read-only SQL and vector database; paste that prompt into Claude Code to give it access to the corpus. 2) Search for phrases relevant to the project, such as KKLT flux compactification or genetic algorithm physics optimization, and use the available text search and SQL filters to inspect matching papers and metadata. 3) Organize relevant results by topic and download the LaTeX sources for key papers. In this run, Claude extracted one paper's source to inspect its MATLAB implementation. 4) Use the resulting topic index and source papers as a starting point for further research.
How it went
The author's posted Claude report described an index of more than 30 relevant papers and two relevant earlier papers, one already in the project's codebase.
Worth knowing
The report noted a 100-result limit for alignment.search() and suggested search_exhaustive() when completeness matters. Its research findings are the assistant's report, not independently verified conclusions.