Writing Is the Best Thinking Tool Ever Invented

Writing is the best thinking tool ever invented because the act of putting an idea into a sentence forces you to discover what you do not yet understand. Every other tool — voice memos, mind maps, AI chat, second brains — lets the gap stay hidden. Writing exposes it on the page where you have to fix it.

That sentence is the entire essay, and the rest is annotation. The Princeton physicist Richard Feynman, asked in 1973 whether his notebook was a record of his thinking, corrected the historian Charles Weiner sharply. "I actually did the work on the paper," Feynman said. "It's not a record, not really, it's working. You have to work on paper and this is the paper."¹ Half a century later the cognitive psychologist Linda Flower and her co-author John Hayes proved Feynman right in the lab, demonstrating in their 1981 paper A Cognitive Process Theory of Writing that "writing is best understood as a set of distinctive thinking processes which writers orchestrate or organize during the act of composing."² The tool was always there. The mistake was calling it something else.

The setup — what most people believe about thinking tools

Most people believe thinking happens first and writing happens after. You have an idea, then you record it. You make a decision, then you document it. You learn something, then you take a note. In this view, writing is a transcript service for the mind — useful, but downstream. The premise feels obvious enough that almost no one tests it.

The premise is wrong, and it has been wrong since at least 1926, when the British psychologist Graham Wallas published The Art of Thought and laid out the four-stage model of creativity — preparation, incubation, illumination, verification — that still anchors creativity research a century later.³ Wallas credited the sentence to an anonymous little girl — "The little girl had the making of a poet in her who, being told to be sure of her meaning before she spoke, said, 'How can I know what I think till I see what I say?'" — and E. M. Forster reassigned it a year later, in Aspects of the Novel, to "that old lady in the anecdote who was accused by her nieces of being illogical."⁴ The organizational theorist Karl Weick made it the cornerstone of his 1995 book Sensemaking in Organizations, where it appears in Chapter Two as the recipe for sensemaking itself: "saying equates to variation, seeing equates to selection of meaning in what was said, and thinking equates to retention of an interpretation."⁵

The point is not that the quote is old. The point is that the most-cited line in the cognitive science of meaning-making says that thinking happens after speech, not before — and writing is speech with the volume turned all the way up.

The pivot — why every other thinking tool fails

Every other thinking tool fails at exactly the moment you stop. Voice memos let half-formed ideas pass uncontested because the voice keeps moving. Mind maps allow contradictions to coexist on opposite branches. AI chat returns coherent paragraphs whether your prompt was coherent or not. Second-brain systems reward capture over synthesis.

This is not a guess. The Microsoft Research team led by Hao-Ping Lee surveyed 319 knowledge workers in January 2025 and found a clean inverse correlation: the more confident a worker was in a generative AI tool, the less critical thinking they performed on tasks the tool helped with.⁶ Four months later an MIT Media Lab group led by Nataliya Kosmyna ran a higher-bar experiment — three sessions of essay writing under three conditions (no tools, search engine, ChatGPT) with EEG monitoring of 54 participants. The result was unambiguous. "Brain connectivity systematically scaled down with the amount of external support," the authors wrote. The brain-only group showed "the strongest, widest-ranging networks." LLM users "consistently underperformed at neural, linguistic, and behavioral levels," and could not accurately quote the essays they had just produced.⁷

The Norwegian University of Science and Technology found something parallel for the keyboard itself. In a 256-channel EEG study published in Frontiers in Psychology in January 2024, Audrey van der Meer and F. R. Van der Weel showed that handwriting "leads to widespread brain connectivity" across parietal and central regions while typewriting does not.⁸ The act of forming a letter — slow, irregular, motorically committed — recruits a different brain than the act of pressing a key. And the act of writing a sentence — slow, irregular, semantically committed — recruits a different brain than the act of speaking one.

Joan Westenberg pushed this argument from the user's side. In her June 2025 essay I Deleted My Second Brain⁹ — which reached 598 points and 348 comments on Hacker News¹⁰, and whose thesis we take apart at length in you don't need a second brain — she described deleting every note in Obsidian, every Apple Note synced since 2015, and "every to-do list from every productivity system I'd ever borrowed, broken, or bastardized." The scale, from her own headline: 10,000 notes, seven years of ideas. Her diagnosis was direct. "Instead of accelerating my thinking," Westenberg wrote, "it began to replace it. Instead of aiding memory, it froze my curiosity into static categories."¹¹ The system did everything except write.

The argument — what the research actually shows

The research shows that writing improves thinking through three mechanisms no other tool offers: forced linearization, forced specificity, and forced re-reading. Linearization makes you sequence ideas. Specificity makes you pick exact words. Re-reading lets your second self catch what your first self missed. Together, they are the engine of clear thought.

Linda Flower and John Hayes's Cognitive Process Theory of Writing (December 1981) used think-aloud protocols to capture, sentence by sentence, what writers actually do as they compose. Their three sub-processes — Planning, Translating, Reviewing — turned out not to be sequential stages but a "hierarchical, highly embedded organization in which any given process can be embedded within any other."² Janet Emig had argued five years earlier in Writing as a Mode of Learning (College Composition and Communication, May 1977) that the unique pedagogical value of writing came from "a cluster of attributes that correspond uniquely to certain powerful learning strategies" — among them the durability of the trace, the pace that matches reflection, and the demand for explicit relations between ideas.¹² Emig's paper became the philosophical backbone of the Writing-Across-the-Curriculum movement that now runs in some form at every research university in the United States.

The empirical case has only strengthened since. Pam Mueller and Daniel Oppenheimer's 2014 Psychological Science paper "The Pen Is Mightier Than the Keyboard" showed that students who took notes by hand outperformed laptop note-takers on conceptual questions; the authors put the gap down to "laptop note takers' tendency to transcribe lectures verbatim rather than processing information and reframing it in their own words."¹³ Henry Roediger and Jeffrey Karpicke's testing-effect studies (2006) demonstrated that retrieval — including written retrieval — produces "substantially greater retention than studying" in delayed tests.¹⁴ Robert Bjork's framework of "desirable difficulties" gives the underlying principle: conditions that slow apparent learning often optimize long-term retention, and writing imposes exactly the right kind of slow.¹⁵

Then there is James Pennebaker. Over four decades and more than a hundred studies, the University of Texas social psychologist has shown that expressive writing — fifteen minutes a day for four days, about something difficult — produces measurable improvements in immune function, blood pressure, and mental health, with an average effect size around d = 0.16 across the literature.¹⁶ The mechanism, Pennebaker has argued repeatedly, is not catharsis. It is the act of "translating experiences into language," which forces the brain to organize what was previously a swarm.

The same conclusion has been reached independently across a thousand years of practice. Niklas Luhmann, the German sociologist whose publication list ran to more than 500 titles by the time of his death, attributed his output to a Zettelkasten of approximately 90,000 handwritten cards¹⁷ that he treated as a communication partner — "As a result of extensive work with this technique a kind of secondary memory will arise, an alter ego with who we can constantly communicate."¹⁸ Vannevar Bush sketched the same insight in As We May Think (The Atlantic, July 1945) when he proposed the memex — a desk-sized associative engine built on his objection to indexed, hierarchical filing: "The human mind does not work that way. It operates by association."¹⁹ Joan Didion put it in autobiographical register in Why I Write (1976): "I write entirely to find out what I'm thinking, what I'm looking at, what I see and what it means."²⁰ Susan Sontag used the same verb in her journals: "I write to define myself — an act of self-creation — part of the process of becoming."²¹ William Zinsser made the case to college teachers in Writing to Learn (1988): "Writing organizes and clarifies our thoughts. … Writing enables us to find out what we know — and what we don't know — about whatever we're trying to learn."²² George Orwell, in Politics and the English Language (Horizon, April 1946), made the political case: "If thought corrupts language, language can also corrupt thought."²³ The cartoonist Dick Guindon — often misattributed to the Turing Award winner Leslie Lamport — gave us the one-liner: "Writing is nature's way of letting you know how sloppy your thinking is."²⁴

The voice that has carried this argument furthest in the modern essayistic register belongs to Paul Graham. His February 2022 essay Putting Ideas into Words states the case in two sentences any writer can recite from memory: "Writing about something, even something you know well, usually shows you that you didn't know it as well as you thought." And: "Half the ideas that end up in an essay will be ones you thought of while you were writing it."²⁵ Graham states the same claim negatively: "if you never subject your ideas to that test, you'll not only never have fully formed ideas, but also never realize it."²⁵ Shane Parrish put his own formulation of it in front of the 600,000-plus readers of his Brain Food newsletter in November 2023, quoting the Farnam Street essay Writing to Think: "Writing is the process by which you realize that you do not understand what you are talking about."²⁶ The most recent voice in the chorus is the editorial from Nature Reviews Bioengineering dated July 2025, simply titled Writing Is Thinking, which argued that "writing compels us to think — not in the chaotic, non-linear way our minds typically wander, but in a structured, intentional manner."²⁷

The danger — what is about to be lost

The thing the next decade will lose is the friction that produced the thinking. The Microsoft and MIT studies are the first wave of a literature that will grow. Workers who let an LLM draft a first sentence rarely write a better second one. The cost is not the essay. The cost is the thought the essay would have produced.

Students who let ChatGPT compose their essays show, on EEG, "the weakest overall coupling" of brain regions across the three test conditions.⁷ The trap has a name in the personal-knowledge-management literature. Tiago Forte calls his system Building a Second Brain — a four-step CODE method (Capture, Organize, Distill, Express) layered on a four-folder PARA structure.²⁸ The system has helped many writers. It has also produced a backlash — Sasha Chapin's essay Notes Against Note-Taking Systems (January 2022)²⁹ and the 478-point Hacker News thread Tired of note-taking apps (July 2020)³⁰ — whose diagnosis is the same across both: capture without writing is hoarding, and hoarding feels like progress.

Steph Ango, the CEO of Obsidian, named the deeper version of the problem in his July 2023 essay File over app: "In the fullness of time, the files you create are more important than the tools you use to create them. Apps are ephemeral, but your files have a chance to last."³¹ The corollary follows by direct substitution: in the fullness of time, the thinking you do is more important than the system you organize it inside, and the system is ephemeral but the thinking has a chance to last — if it was written down in your own sentences.

The dual diagnosis from Andy Matuschak's notes site puts it most economically: "'Better note-taking' misses the point; what matters is 'better thinking.'"³² Matuschak's parent thesis is that evergreen notes are the unit of knowledge work because they are written, atomic, and concept-oriented — not because they are tagged, linked, or stored in any particular app.

The practice — what to do tomorrow

What to do tomorrow is small enough to start before lunch. Open one document. Write one paragraph in your own sentences about a question you have been avoiding. Do not outline. Do not bullet. Do not hand the prompt to an LLM. Read what you wrote. Notice the gap that was hidden until you saw the words. Fix the gap.

The research-backed minimum viable practice is roughly:

1. One question per session. Pick a single question you cannot yet answer cleanly. The Larry McEnerney Craft of Writing Effectively lecture from the University of Chicago (9.2 million views on YouTube as of September 2026) is built on the observation that academic writers waste years writing about topics instead of arguing for claims.³³ A claim is a sentence that could be wrong. Topics are not.
2. Fifteen to thirty minutes of continuous prose. This is the Pennebaker minimum that produces measurable health benefits across his hundred-plus studies.¹⁶ It is also long enough for Wallas's "incubation" stage to give way to "illumination" inside the writing itself.³
3. One re-read in cold blood. Paul Graham's heuristic: "You have to pretend to be a neutral reader who knows nothing of what's in your head, only what you wrote."²⁵ Most of the cognitive value lives here.
4. Cut what was vague. Stephen King's 10% rule from On Writing (2000) — "second draft = first draft minus 10%" — works because the cut sentences are usually the ones that hid the fact that you had nothing to say.³⁴
5. Save the file in a format you will still own in 2046. Plain text. Markdown. Anything you can read in any editor without an account. Steph Ango's argument again: the file is the asset.³¹

That is the entire system. Five steps, one paragraph at a time, every day.

Frequently Asked Questions

Is writing really the best thinking tool, or just one of many?

Writing is the best thinking tool because it is the only one that simultaneously externalizes, linearizes, and preserves a thought in a form you can re-read. Speaking externalizes but does not preserve. Mind maps preserve but do not linearize. AI chat does all three but the thinking is the model's, not yours. Writing is alone in the intersection.

What about people who think out loud or "talk to think"?

People who think out loud often do their best thinking later, when they write down what the conversation produced. The 1926 line Graham Wallas recorded — "How can I know what I think till I see what I say?" — applies to speech as well, but the seeing requires a record. Conversations evaporate. Sentences on a page do not. The talking is preparation; the writing is the work.

Does it matter if I write by hand or on a keyboard?

The 2024 Norwegian EEG study by Van der Meer and Van der Weel found that handwriting produced more widespread brain connectivity than typewriting in 36 university students.⁸ Handwriting wins on encoding. Typing wins on speed and editability. For a daily thinking practice, a keyboard you trust beats a pen you avoid. The sentence still has to be written either way.

Will using an AI assistant for writing destroy my ability to think?

The risk is real but conditional. The Microsoft Research survey of 319 knowledge workers (January 2025)⁶ and the MIT Media Lab EEG study of 54 essay writers (June 2025)⁷ both show measurable reductions in cognitive engagement when LLMs draft the prose. If you use AI to draft, you offload the thinking. If you use AI to interrogate a draft you wrote yourself, you keep the thinking and gain a critic. The first habit is dangerous; the second is not.

Why do so many people abandon their note-taking systems?

People abandon systems because capture feels like progress and is not. Joan Westenberg's June 2025 essay (598 points on Hacker News) traces the pattern from the inside: 10,000 notes accumulated, none synthesized, the system becoming a mausoleum rather than a workshop.¹¹ The fix is not a better system. The fix is to write more sentences and store fewer of them.

How is "writing" different from "journaling" or "note-taking"?

Journaling and note-taking are subsets of writing. Journaling tends toward emotional processing; note-taking tends toward information capture. Writing-as-thinking is the third mode — closer to drafting an essay nobody will read — and it is the mode that produces the cognitive gains documented in Flower and Hayes (1981), Emig (1977), and Pennebaker.^{2 12 16} All three modes produce sentences. Only the third forces argument.

What is the minimum viable writing practice for someone who has never tried this?

One paragraph. One question. Fifteen minutes. Once a day for thirty days. The paragraph does not have to be good. The question does not have to be important. The fifteen minutes do not have to be uninterrupted. After thirty days, re-read the first paragraph. The distance between day one and day thirty is the entire argument of this essay, made personal.

If writing is thinking, then the most important question about any tool you use to think with is whether the thinking you do inside it will still be readable to you ten years from now, in your own sentences. MNMNOTE is a browser-based markdown editor built around that constraint — your notes stay on your own device, works offline, no account, no server, every keystroke producing a plain markdown file you will be able to read on a computer from 2046. Open the tab. Write the paragraph.