TL;DR
Get the little things that make your day delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A November 2026 blog essay by computer scientist Murat Buffalo argues that writing is not just a way to record finished thoughts: drafting can help people understand complex material, make connections and remember ideas. He says AI-generated prose can look like evidence of thought without doing a person’s thinking for them, and recommends starting with an unrestricted draft.
Computer scientist Murat Buffalo has published an essay arguing that writing can function as a tool for thinking, not merely a record of conclusions. He says drafting helps people work through difficult material, make connections and build a memory they can revisit, a distinction he considers more important as large language models produce fluent prose.
Buffalo traces the idea to his own experience as a student. In middle school, he struggled to retain long passages, so he began making extensive notes and mind maps. He says arranging information into branches helped him make sense of what he read. The notes were useful less as a storage system than as a way to process the material.
He later connected that experience to advice from computer scientist Manuel Blum, who compared a finite automaton with a Turing machine. In the analogy Buffalo recounts, the Turing machine’s paper and pencil give it a way to store intermediate results and carry out more complex computations. Buffalo uses the comparison to describe how writing lets people move information outside their immediate working memory. It is an analogy about the role of external memory, not a claim that human thought is literally a computing machine.
Buffalo says he sees the effect when reviewing research papers: he may believe he understands a paper while reading it, then discover gaps when he tries to explain and assess it in writing. The act of composing a review, he says, helps him place the work in context and sometimes notice connections. He also describes a 2018 blog post on similarities between Paxos and Nakamoto consensus that developed into a research-paper idea as he wrote. These examples are his account of how writing has shaped his own work.
Writing Makes Thinking Visible
The essay addresses a practical question for readers who use AI writing tools: whether a convincing piece of text means the underlying reasoning has been done. Buffalo argues that it does not. A person can prompt a model for a polished explanation or memo without first clarifying their own assumptions, while writing a rough draft can expose what they do and do not understand.
That distinction matters in research, study and workplace decisions, where people need to evaluate evidence rather than simply produce readable documents. Buffalo’s argument is that writing can serve as an active test of understanding: gaps become visible when a person has to explain an idea, and a draft gives them something concrete to revise. This is his interpretation, rather than a measured finding about the effects of writing or AI.
He also presents writing as a form of external memory. Buffalo says he returns to his own blog posts to reconstruct ideas he might otherwise forget. The broader implication is that notes and drafts can preserve a trail of reasoning for later use, even when the original understanding was incomplete.
Products Worth Considering
Philips Hue Festavia 65-Foot Smart LED String Lights, 250 Color-Changing Mini LEDs
Blueair Blue Pure 511i Max Smart Air Purifier
Samsung 65-Inch QN70FA Neo QLED 4K Smart TV with 8-Foot WireLogic Ruby Ultra High Speed HDMI Cable (2025)
As an affiliate, we earn on qualifying purchases.
From Mind Maps to AI
Buffalo’s essay moves from a personal study habit to a technical metaphor and then to a concern about AI-generated text. His account begins with mind maps he made as a student in the late 1980s, when Tony Buzan’s approach had reached Turkey. He says the process of deciding where information belonged helped him understand it.
The essay then invokes a statement Buffalo attributes to Leslie Lamport: “If you think without writing, you only think you’re thinking.” Buffalo’s own examples expand on that idea through research reviews, freewriting and notes kept during a project. He recommends treating early writing as a working record rather than a finished product.
He also refers to the six-page narrative memo associated with Amazon and quotes Jeff Bezos’s view that writing such a memo requires clear thinking. Buffalo argues that generative AI complicates that assumption because a model can produce a long, coherent-looking document without the writer having worked through the ideas. The essay does not assess Amazon’s current memo practices or present comparative research on AI-assisted writing.
““If you think without writing, you only think you’re thinking.””
— Leslie Lamport, quoted by Buffalo
What the Essay Does Not Establish
Buffalo’s article is a personal essay, supported by examples from his study and research experience. It does not report a study measuring whether writing improves comprehension, memory or research outcomes, nor does it compare handwritten notes, typed drafts and AI-assisted writing under controlled conditions.
His claims about language models are also expressed as judgments and predictions. Buffalo says he does not trust current models to make creative connections on a person’s behalf, but the essay does not define a test for that ability or establish what models can and cannot do across different tasks. The source does not provide data on how often AI-written documents are produced without substantial human reasoning.
Start With an Unfiltered Draft
Buffalo’s practical advice is to begin by writing freely. He calls this first rule “Mess up”: produce material without stopping to correct typos, polish sentences or judge whether every thought belongs. The aim is to create a rough first draft that can be examined and improved later, rather than waiting for a fully formed argument before writing anything.
He suggests using a running notes file as a kind of lab notebook while doing research or other work. In that early stage, tangents and unfinished thoughts are acceptable; the notes need not be suitable for an audience. The next step is to return to the material, identify weak points and shape the useful ideas into a clearer account.
The essay does not set out a formal method for deciding when AI assistance is appropriate. Its central recommendation is narrower: write and think through an idea yourself before handing the task to a model, particularly when the goal is to discover what you believe or understand. Buffalo’s argument is that generated prose can assist with text, but should not be mistaken for the writer’s own reasoning.
Key Questions
What does Buffalo mean by writing as a thinking tool?
He means that writing can help a person develop an idea while recording it. Drafting forces the writer to organize information, notice gaps and make connections that may not have been clear during reading alone.
Does the essay show that writing improves memory?
No controlled evidence is presented. Buffalo describes his own experience of using blog posts and notes as an external memory that helps him reconstruct ideas later.
What is Buffalo’s concern about AI writing tools?
He argues that a language model can generate polished text without doing the user’s thinking. He warns that asking for an answer immediately may prevent a person from working through their own incomplete ideas, but the essay does not provide a formal test of model capabilities.
What is the essay’s advice for getting started?
Buffalo recommends producing a rough draft without trying to make it polished. He suggests saving notes as work progresses, then returning to revise and evaluate them.
Source: muratbuffalo.blogspot.com
