Writing is Thinking: Why AI Shouldn't Write for You

The author contends that writing is an essential thinking process that should not be outsourced to AI. While AI tools are excellent for research and providing feedback on finished work, using them to generate drafts leads to shallow content. True productivity comes from the depth of original ideas, which can only be achieved through the hard work of manual writing.

Key Points

• Writing is a cognitive tool that forces authors to think deeply about problems and communication.
• Using AI to generate content from bullet points results in skipping the essential thinking step.
• Productivity in writing is defined by the impact of ideas, not the quantity of words or speed of production.
• AI is highly effective for pre-writing tasks like data structuring and post-writing tasks like providing feedback.
• The author suggests asking AI questions about a draft rather than asking it to make improvements directly.

Sentiment

The overall sentiment is cautiously supportive of the article's central warning, with substantial debate about practical boundaries. Hacker News generally agrees that AI should not replace the author's thinking, structure, or voice, but many commenters resist an absolute prohibition and prefer a more precise distinction between authorship and constrained assistance.

In Agreement

• Writing forces the author to discover and refine ideas, so outsourcing drafts or structure can bypass the core cognitive work.
• AI is useful for constrained proofreading, grammar checks, or rubric-based review after the writer has already formed the argument.
• Authentic authorial voice matters; readers should be able to trust that presented prose reflects the human writer's own thinking.
• AI-generated prose often feels generic or hollow because it can smooth over weak reasoning and homogenize style.
• Research and translation assistance require verification because fluent model output can conceal errors, bad sourcing, or unintended meaning.

Opposed

• The boundary between research, outlining, proofreading, feedback, and writing is blurry, so the article's rule can seem inconsistent.
• LLMs can be tools like word processors or research assistants; the real issue is low-effort or bad writing, not AI involvement itself.
• Using AI for coding while warning against AI for prose appears inconsistent if both activities require thinking.
• Some commenters argue that careful, iterative use of AI can preserve human intent while expanding editing and transformation capabilities.
• A few reject anti-AI framing as moral panic and argue that slop predates AI, even if AI makes it easier to produce.