In the world of AI, it is just too easy to outsource your writing to an LLM model. Ask Gemini (2.5 pro - my favorite model) to write on an idea you’ve been thinking about and bam - you’ll have a pretty good finished product on your hands.
But what have you really achieved through this? A piece of writing that is mostly fluff to your brain. Why is it fluff? It’s fluff not because of the writing style or content (on the contrary I’ve seen Gemini produce excellent metaphors and analogies with the right prompting), it’s fluff because your brain hasn’t spent any time on figuring out those words. It’s not spent any time figuring out structure, arguments, contradictions, other connections & examples. Your brain has simply read AI’s input. It has not spent time thinking about the idea. When you try to speak about the topic to someone else, you’ll notice the glaring gaps in your understanding of the topic.
“But I’ve ‘written’ about this” - your brain will say. You’re only fooling yourself. You’ve only read about your idea, but you haven’t built the scaffolding around it to truly understand it. This is what Jordan Peterson says about the writing process:
"To summarize what I've learned, I need to iron out the contradictions between what I've learned, and I need to elegantly formulate that. I need to get my word choice right, my phrase choice right, and my sentence choice right. I need to organize the sentences into proper paragraphs, and the paragraphs into proper sequence so that I have a coherent argument. At the same time, what you're doing is integrating your own personality at the highest and most abstract level of organization. You're sharpening your tools, and you're putting yourself straight because you're learning to think. You learn to do that by writing."
"You learn to think by writing. Now, there's more to read, to speak, and all of that, but the best thing you can do is read and write every day. A couple of hours every day, write about things you find important, and see if you can discover what you believe to be true. That'll build you a foundation, and it's unbelievably practical."
If you want to be a better communicator, learn to improve your persuasion skills through writing. When your writing is clear, your thinking is clear and your ideas become powerful. Shane Parrish explores this idea in his book Clear Thinking.
This is not to say that AI shouldn’t be leveraged at all. Personally, I think AI can help writing in the following ways:
1/ Research
AI is an excellent research partner — as long as you double check the sources it’s referencing. To make sure that I don’t break my flow of writing, I’ll add a placeholder where I don’t know the exact dates or stories. After my writing session, I go back to that document and ask AI to help me fill out the background research. This is a massive time save.
2/ Editing
Since I’m actively writing, I want to make sure I’m improving with each piece I put out. After crafting handwritten notes, I ask AI to grade the writing. Pro tip: don’t tell AI that you’re the author to eliminate positive bias. Continue prompting the model to critique and refine your structure.
3/ Admin Writing
Everyone has pure utility writing to do — routine emails, meeting logistics, operational notes. If the outcome of this kind of writing doesn’t improve your thinking, you should export that entirely to AI. As David Perell noted, all forms of pure-utility writing will soon be handled by AI. Keep the deep thinking for yourself.
In summary: develop a writing habit to sharpen your thinking. Don’t outsource the cognitive work of writing to AI. Think of it as going to the gym for your brain. Carefully use AI to give your writing leverage through editing and research, and automate the administrative writing.
I also recommend reading my notes on How to Take Smart Notes and How to Be a Writer. I've also written about how AI can be leveraged for developer workflows in my post on Cursor.
Originally published on Sameer Bajaj's Blog.
Top comments (0)