How Nodus can support research in the humanities

Humanities projects have an awkward habit of refusing to fit in one tool. A historian ends up moving between books, archival scans, photographs, interview recordings, notes and a spreadsheet nobody else could read. A literary scholar juggles editions, marginalia, secondary literature and a slowly thickening web of concepts. Anyone working on genealogy or prosopography is tracking people, places, relationships and a great deal of evidence that doesn't quite agree with itself.

I wrote Nodus because that was my week, most weeks, while organising material for a doctorate. Not because the individual tools were bad. Zotero is excellent. So is a good PDF reader. The problem was that none of them knew about each other, so the connections between them lived in my head, and my head kept losing them.

Nodus is local-first, so everything in a vault sits on your own machine. AI is optional, and when you use it you choose whether it goes to an external provider or to a local model. That matters more in the humanities than people outside it tend to assume, because a lot of humanities material is not yours to upload: an unpublished archive, a family's photographs, an interview recorded under a consent form that says exactly where the audio may go.

The thing the humanities need most from research software is unglamorous. It's the ability to say where an interpretation came from. In Nodus, a finding stays attached to the passage, page, image or timestamp it was drawn from, which means that six months later you can check yourself instead of trusting your memory.

Working from a research library rather than a folder of PDFs

The Academic vault is where most humanities projects start. It can hold a library you build inside Nodus, or read the Zotero collection you already keep. Either way the sources stop being isolated bibliographic records and become part of a network of claims, findings, methods, notes and ideas.

A literature review is the ordinary case. You import a few dozen papers, or point Nodus at Zotero, and read inside the app. Passages you care about get attached to ideas. As more sources come in, the relations between them get easier to see, semantic search finds material across the whole library even when the wording differs, and the graph shows you what is connected to what.

That helps most when the research question is still moving, which in the humanities is most of the time. Instead of opening a new document every time something occurs to you, the idea goes in attached to the sources that support it and the ones that argue with it. After a year of that, the shape of the literature is visible: where it's crowded, where two camps are talking past each other, where your question has almost nothing standing behind it.

For writing, the point is that the reading is still there when you need it. Material gathered during the review is available while you draft, and citations lead back to the evidence, not to a hazy recollection of which chapter it was in. There's a writing environment inside Nodus, plus copilots for Word and LibreOffice, so the research and the drafting don't have to happen in separate universes.

Historical work has a specific version of this problem. If you're reading two hundred books looking for how a concept, a place or a representation shifts between authors and decades, finding another instance of the word is the easy part. What you actually need is to land back in the paragraph and see the context, because in the humanities the context usually is the finding.

Working with archival sources, people, testimony and structured historical data

Plenty of humanities research never starts from a conventional academic library at all. It starts with a box of documents, a folder of photographs, a set of recordings, a catalogue somebody kept by hand in 1954. Nodus has separate vault modes for those, because pretending an archival photograph is just another PDF makes everything downstream worse.

Primary Sources is for working critically with documents and manuscripts. OCR and transcription sit beside the original image, and the layout keeps the distinction that matters: what the document says, versus what you have concluded from it. With handwriting, bad scans, or a text whose meaning depends on where it was filed, blurring those two is how errors get laundered into certainty.

Testimony is built around interviews and oral history. Transcription runs locally, speakers are separated, and consent and access conditions are part of the record itself, not a note you will fail to find later. Passages keep clickable timestamps, so an analytical claim is one click from the moment in the audio where somebody actually said it. If you have ever tried to relocate a thirty-second answer inside a two-hour recording, you know why that's the feature.

Genealogy takes the same workspace and points it at reconstructing families and individual lives from primary sources, with a tree, a timeline and a map, and GEDCOM in both directions. Events keep the evidence behind them. Genealogy gets filed away from academic history more often than it deserves, because its hard problems are the historian's hard problems: identities that might be the same person, records that disagree, gaps that can't be closed, and the permanent obligation to keep what is documented apart from what is inferred.

Prosopography is the same question at population scale. Rather than one life, you're following a group across scattered and inconsistent sources. Name variants and uncertain identities get handled without being flattened into a single confident answer, which is the whole methodological point. It suits studies of political groups, intellectual networks, institutions, professional bodies and other historical populations where the individuals only matter in aggregate.

Databases covers the part of a humanities project that has quietly become a dataset. Typed fields, relations, formulas, attachments, reusable filtered views, CSV import for whatever you already built in a spreadsheet, and analysis over your actual rows. For a historian that might be a catalogue of photographs, a database of travellers, an inventory of archival files, or a table of places and events. The advantage is not the table. It's that the table lives next to the books.

Connecting qualitative interpretation with quantitative and computational work

Very few humanities projects stay purely qualitative or purely quantitative, whatever the funding application said. You start by reading, and then you find yourself building a table to compare authors, dates, places or themes. Close reading turns into a corpus. Interpreting forty photographs turns into structured information about four thousand.

Nodus holds both halves in the same workspace. The research library keeps the sources and the interpretation; a Databases vault keeps the structured information drawn from them. Neither replaces the other, and they answer different questions about the same project.

AI shows up at different points. In a library it can extract ideas, propose relations, surface contradictions and flag where the evidence thins out. In a database, AI columns can classify rows while staying inside the fields and values you defined, so it can't invent a category that isn't in your data. With Ollama or LM Studio, all of that can run offline.

None of it replaces your judgement, and I'd be wary of anyone selling it that way. A generated finding is a hypothesis until you've checked the source. A proposed relation between two documents is a suggestion, not a fact. A classification produced by a model is not a historical category just because it's consistent. Keeping the evidence attached to the finding is what makes that scepticism cheap to act on: checking is one click, so you actually do it.

Here's a shape I see often. Someone is working with several thousand photographs from a regional archive. There's a Databases vault holding structured information about each image, a Primary Sources workflow for reading the originals, notes recording what they think is going on, and an Academic vault with the secondary literature that gives it context. AI does the repetitive parts. The interpretation stays where it belongs.

The same is true for a corpus of books. Bibliography in Zotero. Reading and passage-marking in Nodus. Recurring ideas connected back to specific sources. Structured data for the comparisons that need counting. What you end up with isn't a stack of AI summaries. It's a workspace where the sources, the interpretations and the numbers still know about each other.

Choosing a Nodus workflow for different humanities projects

There is no single humanities method, so Nodus doesn't impose a single project structure. Academic is the general-purpose starting point when the project revolves around a bibliography, documents, ideas and scholarly writing. Databases is for projects that lean on structured information. Primary Sources fits archival documents and manuscripts, and Testimony fits oral history and interviews.

Genealogy and Prosopography come into their own when people and relationships are the object of study and not just the backdrop, and both combine happily with the rest. A study of an intellectual network can put prosopographical data about the individuals alongside an Academic library of the scholarship and the primary sources. A project about one family can put genealogical records next to documents, photographs and the contextual literature.

Teaching and Study are in the same application, which matters if you teach as well as research. Teaching organises courses, materials, activities and assessment; Study handles subjects, materials, notes, calendars and revision. You can use one, both, or neither without the others getting in the way.

So the question to ask about Nodus isn't which feature replaces a tool you already have. It's which parts of your workflow are currently disconnected and shouldn't be. Keep Zotero for the bibliography. Keep your recorder and your transcription habits if they work. Keep the spreadsheet if the spreadsheet is fine. Nodus is meant to sit underneath those and hold the connective tissue: the sources, the evidence, the ideas and the structured data that together make up the project.

That's the part I'd defend hardest. Nodus doesn't decide what an interpretation should be, and it will not turn evidence into conclusions while you have lunch. What it does is keep the path from source to interpretation visible, whether the source is a book, a parish register, a photograph, an interview or a row in a table, so that when the argument changes, and it will, you can find your way back to the material that started it.

The academic research guide explains the source-to-evidence workflow in detail. Researchers who already collect in Zotero can also compare the broader Zotero integration with the focused Nodus Zotero plugin.