The author juxtaposes traditional note‑taking with a modern AI chat, highlighting the tension between external memory and machine assistance.
WRITING AS A THINKING TOOL REDEFINED BY AI
*Writing transforms a mind from a finite automaton to a Turing machine, the author argues. The rise of LLMs threatens the habit, but the practice still offers a measurable cognitive edge.*
In a November 07, 2026 blog post, a former middle‑school struggler recounts how note‑taking rescued him from unreadable pages. The late‑80s arrival of Tony Buzan’s mind maps turned his desperation into spider‑web notebooks that processed, not stored, information. A decade later, a PhD‑level procrastination binge led him to Manuel Blum’s stark comparison: without paper, a thinker is a finite automaton; with paper, a Turing machine. The claim reframes writing as a computational upgrade, not a mere record‑keeping habit. As large language models flood the market, the author asks whether the external memory advantage of writing still matters, or if AI will render the practice redundant.
Why does writing upgrade a mind from finite automaton to Turing machine?
The author cites Manuel Blum’s graduate‑student advice: finite automata can add but not multiply; Turing machines can compute any computable function. The only extra ingredient a Turing machine has is "paper and pencil"—the act of writing. By externalizing thoughts onto a medium, a person moves from the limited state‑space of a finite automaton to the unbounded computational power of a Turing machine. The claim is crisp: remove the paper and you downgrade an entire complexity class. The author links this leap to historic milestones, noting that language externalized thinking, and writing externalized even more, propelling human intelligence forward.
How does freewriting spark research breakthroughs?
The author describes a 2018 blog post that began as a casual note on Paxos and Nakamoto consensus. As the freewriting progressed, the connections sharpened, eventually seeding a research paper. The process mirrors "out‑of‑core execution" in computer science: the writer offloads intermediate reasoning to external memory, freeing mental bandwidth for larger insights. A similar pattern appears when reviewing academic papers; the author often believes they understand a work until the review draft reveals a 60% comprehension gap, which writing then closes. The practice turns tentative notes into a lab notebook, allowing tangents to surface and be evaluated.
Can large language models replace the thinking benefits of writing?
The piece questions whether LLMs make writing‑for‑thinking obsolete. It notes that models have ingested the public literature but lack access to a user's half‑formed intuition. The author warns that opening a chat instantly bypasses the personal drafting stage, eroding the cognitive workout that writing provides. Bezos’s six‑page memo is invoked as proof that structured writing forces clear thought—yet the author argues that LLMs can now generate six pages without any underlying reasoning. The conclusion: LLMs can supply information, but they cannot substitute the mental discipline of putting pen to paper.
What practical steps boost writing‑for‑thinking?
Rule 1, per the author, is to "mess up"—produce bulk text without self‑editing. The goal is a zeroth draft that the mind can later critique. The author references McPhee’s argument that any written output, even a flawed session, gives the brain material to react against. The advice continues: treat a notes.org file as a lab notebook, write while working, explore every tangent, and avoid pressure to be presentable. This raw accumulation creates the external memory the author describes as essential for rapid recall and deeper understanding.
Frequently asked questions
Why does the author compare writing to a Turing machine?
Manuel Blum’s advice says a Turing machine’s only extra resource over a finite automaton is paper and pencil, implying that writing upgrades a mind’s computational power.
How did freewriting lead to a research paper in 2018?
The author started a blog post on Paxos and Nakamoto consensus, let the connections emerge while writing, and the post became the seed for a later research paper.
Can large language models replace the thinking benefits of writing?
The author argues LLMs lack access to a user's personal intuition and cannot force the mental discipline that writing provides, so they cannot fully replace writing‑for‑thinking.
The evidence is clear: writing externalizes thought, expands cognitive capacity, and fuels breakthroughs that AI alone cannot generate. As LLMs grow louder, the discipline of drafting remains a rare comparative advantage. Those who keep the pen in hand will continue to outthink the models that merely echo existing knowledge. The future of intelligent work hinges on preserving that manual edge.
Sources: Hacker News – Writing as a thinking tool (http://muratbuffalo.blogspot.com/2026/11/writing-as-thinking-tool.html)