Claude Code for writers

A draft is only as good as the research trail underneath it, and that trail almost never survives contact with a deadline. Here is what changes when the reading and the writing share the same filing system instead of living in two different places.

Writing that draws on research has a two-part failure mode. The reading happens in one place, scattered across articles, PDFs, interview notes and half-remembered conversations. The writing happens somewhere else entirely, usually a blank document, at which point you are reconstructing from memory which source actually said the thing you are about to attribute to it.

The gap between those two places is where good writing gets slower than it needs to be. Not because the research was thin, but because finding it again, three weeks after you read it, costs more time than doing the reading did in the first place.

There is a compounding effect that only shows up over months. A writer who has been filing sources carefully for a year is not just faster on any single piece, they have accumulated a body of connected reading that new work can draw on directly. A piece written this month can reference and build on research done for a piece six months ago, because the earlier material was filed somewhere it can actually be found and reused, rather than living only inside the finished article it originally supported.

Keeping the source attached to the claim

The fix is capturing what you read at the moment you read it, filed against the piece it is actually for, rather than a general someday-pile. That is the specific habit behind save articles for AI: a reading capture workflow, and it means that by the time you sit down to write, the sourcing work is already mostly done rather than still ahead of you.

The connective tissue between ideas matters as much as the individual sources. A piece that draws on a dozen different reads is really an argument about how those reads relate to each other, and building that web of connections as you go, rather than trying to reconstruct it under deadline, is the approach covered in Zettelkasten with AI: automatic linking.

Where this differs from a research tool that only summarizes

Tools built to summarize a document you hand them are useful for a single read, but they do not build the connective structure across everything you have read over months, which is the actual asset a working writer needs. That trade-off is covered directly in NotebookLM vs Obsidian for research notes, and it is worth understanding before choosing a tool based on how good its summaries look in a demo.

If your writing is closer to academic or technical research than to journalism or essays, the same filing discipline is covered from that angle on the researchers page, and the two roles overlap more than the job titles suggest.

Revising without losing the trail

A piece rarely survives its first draft unchanged. Sections get cut, arguments get reordered, and a source that anchored an early paragraph sometimes ends up supporting a completely different point by the third revision. If the source is only loosely attached to the draft, that kind of restructuring quietly breaks the link, and you discover it during fact-checking, at the worst possible time. Keeping the source filed against the claim rather than against the document means the connection survives the revision, because it was never tied to where the sentence happened to sit on a given day.

Fact-checking as a byproduct, not a separate pass

Fact-checking under deadline is miserable precisely because it usually happens after the fact, as a separate pass through a finished draft, chasing down sources that felt solid in the moment and now need to be re-verified. When the source is filed against the claim from the start, the fact-check is mostly already done by the time an editor asks for it, because the connection was never allowed to go stale in the first place.

Getting started

Start with your current piece. File every source as you read it, attached to that piece specifically, and see whether the draft goes faster because the sourcing work is already done by the time you start writing.