A research workflow that still works when you sit down to write

Browser folders named “thesis” fill up because collection feels like progress. Writing exposes the lie: half the links are background you already know, a quarter are behind logins you no longer have, and three of them are the same paper saved under slightly different titles. A research directory should be smaller than your curiosity. It should be big enough to support the argument you are actually making.

Capture with a job, not a vibe

When you save a source, pick one role for it. We use three:

• Evidence — it supports a specific claim. Your note should name the claim.
• Background — it helps you understand the landscape. You probably will not cite it.
• Example — a method, a figure, a comparable case. Useful, but not proof.

If you cannot pick a role, you are bookmarking anxiety. That is allowed once in a while. It is a bad default. Put maybes in a separate directory called “maybe” and do not mix them with sources you would defend in a seminar.

Write the citation while the tab is open

Author or organization, year, and the page or section if it matters. Then the reason: “Table 2, household energy use, used in the methods comparison.” Future-you cannot reconstruct that from the domain name. If the source has no identifiable author, treat that as a quality flag. You can still keep it as background. You should think twice before building a paragraph on it.

Titles on academic pages are often unreadable. Replace them with something you would search: “Muller 2021 — survey of remote onboarding in SMEs,” not the 90-character journal title in Title Case.

Tags as an index, not a folksonomy

Keep a short list and reuse it across papers:

• topic: climate, labor, education
• format: paper, dataset, report, news
• status: cite, background, outdated

Status is the tag people skip. Marking something outdatedis kinder than deleting it if you still need to show you looked. It also stops you from citing a 2014 blog post as if it were current practice.

Split when the pile gets rhetorical

One directory per paper or unit of work. If you are writing two chapters, two directories. “All research” is how you lose an afternoon. A second directory called “sources to cite” versus “background reading” is a useful split even inside one project.

Share a research list publicly only if every item is something you have the right to point to and you are comfortable with strangers treating it as a recommendation. Course reading lists can be public. Your unpublished interviews cannot. When unsure, keep it private and invite the supervisor.

Setup steps are in Getting started. The weekly pass that catches dead library links is in the weekly review.