Writing as a thinking tool

Writing—whether longhand on grid paper or in a plain text file—is framed as a primary tool for thinking, designing, and truly understanding problems, not just a way to record or polish finished ideas. Commenters describe how drafting prose, sketches, and code surfaces hidden assumptions, clarifies what they are “actually” building or researching, and creates durable context that AI-generated text or code often fails to provide. While many see value in LLMs as editors, rubber ducks, or research helpers, they warn that outsourcing the core act of writing risks shallow understanding, weaker learning, and the loss of important cognitive benefits that come from grappling with ideas yourself.

Pen-and-paper as a thinking tool

  • Many describe projects starting as handwritten scribbles, boxes, and arrows that “unlock” clarity before any code.
  • The freedom and lack of friction of pen and paper are repeatedly preferred over tablets or digital tools.
  • Grid paper (especially 5mm) and simple, reliable pens are favored; some use color-coding systems for services, infrastructure, and priorities.
  • Several workflows combine analog capture with digital archival (photos → OCR → Obsidian/markdown).

Writing and clarity of thought

  • Writing exposes unclear thinking; complex prose is seen as a frequent cover for sloppy or evasive reasoning.
  • Multiple commenters note discovering what they were actually researching or building only while writing.
  • Drafting plans in natural language before coding surfaces edge cases and design decisions that were invisible in the head.
  • Not all writing should be published; much is “thinking exhaust” that serves only the author.

AI’s role in writing and thinking

  • Strong concern that AI-written text short-circuits the thinking process; AI output is treated as a heuristic for shallow ideas.
  • Some refuse to engage deeply with content if it appears LLM-generated or if replies are obviously AI-mediated.
  • Others use LLMs as editors or “fast feedback” partners (consistency checks, clarity suggestions), stressing that core drafting must remain human.
  • A minority argue that, used carefully (summarizing, probing assumptions, adversarial review), LLMs can enhance understanding rather than replace it.

Programming, design, and writing as parallel processes

  • Coding is framed as another form of writing for thought: real complexity and missing assumptions emerge only in implementation.
  • There is worry that relying on AI for code generation erodes this learning loop.
  • Some mitigate this by retyping AI-generated diffs, or by writing the behavior in English first; understanding, not typing speed, is seen as the bottleneck.
  • Designers echo this: sketching in Figma or on paper reveals and refines inchoate “thought-feelings” in ways prompting alone does not.

Medium, habits, and limitations

  • Handwriting vs keyboard is debated, but most value low-friction, free-form sketching plus later digital structuring.
  • Daily freewriting, morning pages, and lab-notebook-style logs are reported as powerful for idea generation and tracking complex work.
  • One dissenting view holds that writing can become over-analysis for highly analytical people, turning into pseudo-rationalization rather than genuine progress.