Building a Second Brain for Research: What Survives Contact With a PhD

A second brain for researchers works when you treat Tiago Forte’s system as a parts bin rather than a doctrine: the CODE loop maps cleanly onto research practice, progressive summarization and intermediate packets are worth stealing, and PARA — the organizational scheme at the system’s center — breaks on contact with research’s time horizons. This article sorts the framework into what survives a PhD, what doesn’t, and what it costs to run, ending with the weekly routine that keeps the whole thing from becoming a very organized graveyard.

The context, briefly: Building a Second Brain (Forte, 2022) systematized a decade of his courses into two acronyms. CODE is the workflow — Capture what resonates, Organize for action, Distill to the essence, Express in output. PARA is the filing scheme — every piece of information lives under a Project (short-term, with a deadline), an Area (ongoing responsibility), a Resource (topic of interest), or the Archive (everything inactive). The book was written for knowledge workers generally, and it shows: the load-bearing assumption throughout is that knowledge exists to serve projects, definable efforts with end dates. Researchers have those. They also have something the framework never quite metabolizes — a literature, a set of questions, and an intellectual identity that run on decade timescales and never resolve into a deliverable.

Where PARA breaks

PARA’s core move is organizing by actionability: information sorts by when you’ll act on it, not what it’s about, and things migrate toward the Archive as projects close. For a consultant shipping decks, this is genuinely clever — it keeps the active surface small. For a researcher it misfires in three places.

First, the interesting boundary is unstable. Your dissertation is a project by any definition, but where does “the literature on measurement invariance” live? It serves the current chapter (Project), your standing competence (Area), and your curiosity (Resource) simultaneously, and PARA’s own tie-breaking advice — file it where you’ll act on it — changes answer month to month. A scheme that makes you re-litigate filing decisions is charging rent on every capture.

Second, the Archive is where research materials go to be lost. In PARA, a closed project’s materials archive wholesale. But a researcher’s closed project is not inert — the papers you read for your masters thesis are the foundation of the dissertation; the failed study’s literature review resurfaces in the grant application three years later. Research knowledge doesn’t retire when a deadline passes; it compounds. An actionability-based scheme systematically buries exactly the material whose value is long-horizon, which is to say: the research.

Third, notes filed in folders can’t serve two masters; notes in a network can. This is the deeper issue. A claim-note about loss aversion may need to appear in a chapter outline, a grant argument, and a lecture — simultaneously, and for years. In a folder scheme it lives in one place and gets forgotten in the others; in a link-organized vault it sits in three maps of content at once and surfaces from any of them. The resolution isn’t to reject organization; it’s to apply PARA where it fits and links where they fit, which is exactly how the pieces map below.

None of this is a takedown — Forte was solving a different problem, for people whose knowledge serves deliverables rather than constituting the work itself. Diagnosing the mismatch precisely is what lets you keep the parts that transfer.

CODE, mapped honestly onto research practice

The four-step loop survives translation almost perfectly — provided each step lands on the right tool.

Capture = Zotero. Forte’s capture advice (“capture what resonates”) is aimed at people saving articles and podcast quotes into a notes app. A researcher’s capture layer already exists and is better: the reference manager owns papers, PDFs, metadata, and — if you annotate in its reader — your highlights and marginal comments. The one Forte-ism to import is restraint: capture is cheap and processing is not, so the browser-connector click is a commitment you’re making on behalf of your future processing time. Capturing forty papers a week you’ll never process isn’t a second brain; it’s a backlog with a database.

Organize = the numbered vault. Notes live in Obsidian, in a shallow numbered folder structure (literature notes in one folder, permanent notes in another, projects, attachments — the layout detailed in the Obsidian for academic research guide). Here PARA earns a partial rehabilitation: it’s genuinely good for the project folders — active paper drafts, the current grant, teaching this term — where deadline-scoped containers match reality. Use PARA thinking for project materials; use links for knowledge. The failure mode is applying it to both.

Distill = literature notes. Forte’s distillation tool is progressive summarization: bold the best passages of a captured article, then highlight the best of the bold, layering compression each time you revisit. The researcher’s version is stronger and predates it — the literature note: claim, evidence, method, limitation, connections, written in your own words within 48 hours of reading. Progressive summarization is worth keeping in moderation for one case: long sources you’ll revisit repeatedly (a methods textbook, a landmark review), where layered bolding on the extracted annotations genuinely speeds the fifth re-read. Applied to every capture, it’s a procrastination format — three passes of decorating other people’s sentences instead of one pass of writing your own.

Express = drafts and MOCs. Forte’s best idea lives here: knowledge work should ship intermediate packets — small, reusable units of finished thinking — rather than accumulating toward some someday-magnum-opus. Academia already runs on intermediate packets without naming them: the conference talk that pilots a chapter, the methods appendix reused across three papers, the two-paragraph literature synthesis that becomes a review’s introduction, the map of content that is secretly a paper outline. The practical upgrade is deliberateness — when you finish a good MOC or a tight literature synthesis, recognize it as a packet: it’s a lab-meeting presentation, a Twitter-thread-turned-blog-post, a conference abstract. Expressing early and small is how a note system pays dividends during the PhD instead of only at the defense.

What to steal, what to skip

Steal five things. The capture-restraint mindset — resonance as a filter, applied at the browser connector, where over-capture is cheapest to prevent. Intermediate packets, named and hunted deliberately; once you have the concept, you’ll notice you were already producing them and throwing half of them away. Progressive summarization, but only for the five sources you’ll read five times. PARA for project folders and only project folders. And Forte’s genuinely correct meta-point, which academics need more than anyone: organization exists to serve output. A note system that isn’t producing drafts is a collection, whatever it cost you, and “I’m still setting up my system” is a sentence that should frighten you by its third month.

Skip five things. PARA as the scheme for your notes — links do that job, for the reasons above. The Archive as a destination for anything literature-shaped; a researcher’s archive should hold dead admin, not dormant knowledge. Capture-everything maximalism: a read-later queue of two hundred items is anxiety with a progress bar, and the correct response to most of it is deletion without guilt. The app tour the second-brain internet runs on — Notion this quarter, Tana the next, Capacities after that. (If you’re genuinely undecided rather than touring, we settled the Notion-versus-Obsidian question separately.) The tools for this workflow are settled (Zotero plus Obsidian, both free, both file-based, both older than most of their competitors’ funding rounds), and every migration costs a month of momentum plus some percentage of your links. And finally, distillation rituals applied indiscriminately, which is how people end up with four layers of highlighting on a paper they never wrote a sentence about — ceremony metabolizing the time that understanding needed.

The maintenance-cost ledger

The second-brain literature is systematically silent about carrying costs, and carrying costs are what kill systems mid-PhD. So run a ledger: every component of the system must pay rent — measurable time saved or output produced — or it gets cut. Daily notes you never re-read: cut. An elaborate tagging taxonomy alongside your links: cut (one system of connection, maintained, beats two half-maintained). Dashboard plugins that query your vault to display statistics about your vault: cut, with prejudice. A weekly review template with eleven sections: cut it to the three you actually answer. The test for any component is brutal and portable: if I stopped doing this, what concretely gets worse, and would I notice within a month? If you can’t name the loss, you’ve found decoration.

This discipline matters because a research second brain must survive the weeks that stress-test it — fieldwork, teaching crunch, the reviewer-2 revision from hell. A system with low idle cost degrades gracefully: skip two weeks and you’re behind on processing. A high-ceremony system degrades catastrophically: skip two weeks and the backlog plus the guilt plus the eleven-section template mean you never come back.

A realistic weekly routine

Forty-five minutes, calendared, once a week. This is the entire maintenance burden of the system described across this site, and each step has a purpose you can name:

1. Process the reading backlog (25 min). Every paper annotated this week gets its literature note, per the 48-hour rule — the weekly block is the backstop for the days that rule slipped. If the backlog exceeds the block two weeks running, the fix is reading less, not processing faster.
2. Promote and link (10 min). Ask of the week’s literature notes: did anything here earn a permanent note? Write the zero-to-three that did, and link them into the relevant maps of content. This is the step that compounds; see the practical Zettelkasten for the mechanics.
3. Touch the active project (5 min). Open the MOC for your current chapter or paper. File anything new that belongs to it. Notice what its outline is still missing — this is where next week’s reading list actually comes from, which closes the loop between the system and the work.
4. Empty the inboxes (5 min). Fleeting notes, the Zotero unfiled folder, the read-later queue. File, delete, or consciously defer. Deleting is a legitimate outcome; most captured things were only interesting once.

Nothing in the routine reviews the system itself. That’s deliberate. The full pipeline this routine maintains — capture through drafts — is laid out in the researcher note-taking system.

What AI actually does for a second brain in 2026

The honest answer, mid-2026: one thing well. Semantic search over your own notes — retrieval by meaning rather than keyword — genuinely fixes a real failure mode, the note you wrote eighteen months ago under vocabulary you no longer use. In Obsidian this now arrives via a maturing plugin ecosystem: Smart Connections surfaces related notes from local embeddings as you write, Copilot-style plugins let you ask questions conversationally against your vault as the corpus, and local-model setups keep the whole thing off the cloud for those with sensitive material — a live concern for anyone holding unpublished data or interview transcripts. I’m mentioning the landscape rather than reviewing it; the plugins are changing quarterly and deserve their own piece.

The caution is the same one this whole article runs on: AI retrieval raises the value of notes worth retrieving, and does nothing for a vault of unprocessed captures — semantic search over forty highlight-dumps returns forty highlight-dumps, faster. And the distillation step is precisely the one you cannot delegate, because writing the claim in your own words is the understanding, not a report of it. AI is a better librarian for your second brain. It is not a substitute for having thought — which was, all along, the only component of the system that couldn’t be skipped.