Roadmap
Where this is, where it is going, and what it will not become. If the gap between what it does now and what you need sits in the first two lists, it is likely to close. If it sits in the last, it will not.
This page is the owner’s current statement of intent, written by hand and kept up to date by hand. The live list of open work, item by item, is on GitLab; where the two disagree, that list is the one to believe.
Built and in use
Every pivotal paper placed by year and sized by how much later work built on it for the term, with the relationships between specific papers typed and named.
Each definition is quoted word for word from the paper, and checked against the text this system pulled out of it. Quotes that cannot be found are marked inferred rather than dropped or quietly smoothed over.
The parallel lines of development a term followed, each readable as an ordered account of the papers that carried it.
What a paper contributed, how it defines the term, what it drew from and what drew from it, with the reasoning behind each link shown as the tool's reading rather than as fact.
A curator can rename, merge, reclassify and remove. Every change is recorded, so a map shows how much human judgment it has had.
Take a map, or a single thread of it, into your draft as a Markdown scaffold with the quotes and the citations attached, plus BibTeX for exactly the papers it cites. The structure and the evidence are real; the argument is yours to write.
Working on next
The single biggest limit on who this is useful to. Finding free copies of papers by their DOI, and reading biomedical full text, would open up fields that do not post preprints to arXiv.
Part of this is already live: on a thread's account, where it has been run, each thing a paper did carries its own quote, matched against that paper's text or plainly marked as unmatched. What is left is doing that for every map and every account, so a whole reading is auditable sentence by sentence rather than paragraph by paragraph.
Point it at a folder of PDFs or a BibTeX file instead of having it choose papers for you. Finding papers and reading them are already separate steps inside the system, so this is more wiring than invention.
Showing the candidate papers, the year span and the cost before a map is built, so you can judge the papers before spending on them.
Under consideration
Today the question each paper is asked is fixed: how do you define this term? Letting a reviewer set that question turns the same machinery into a general extraction tool.
How a term's development differs between fields, or how a map changes as new work lands.
Not planned
This maps how a term developed. It does not rank papers by quality, predict what to read next, or tell you what matters in your subfield.
Finding papers on a topic is well served elsewhere and served better. This starts where search ends: you have a term, and you want its development.
The pipeline drafts; a person decides. Removing the human step would remove the thing that makes a curated map worth more than its raw output.
Built in the open by one person, so this is an honest order of intent rather than a delivery schedule with dates. If something here decides whether the tool is useful to you, request a concept and say so in the note.