Why I Write (and Why You Should Too) - Your Mind’s version control
Every now and then, people ask me why I write. I don’t get paid to write here, and for an engineer trading time for value, the ROI isn’t immediately obvious.
However, I consider writing one of the highest-leverage activities a technical leader can do. Here is why.
1. Writing is Cognitive Resistance Training
Writing isn't just recording thoughts; it is the whetstone that sharpens your intellect. The more you write, the sharper your reasoning becomes. You learn to construct clearer arguments and crisper explanations. _1
Consider the famous "Amazon Six-Page Memo" rule. Jeff Bezos banned PowerPoints because bullet points allow you to hide weak logic. Writing a full narrative forces you to connect every dot. If you can't write a coherent paragraph about a complex architectural decision, you likely don't understand it well enough to code it.
2. Feedback Debugs Your Logic
Writing publicly invites feedback, which exposes weak arguments and strengthens good ones.
I once wrote a post claiming that Matrix Factorization was obsolete in the age of LLM-based embeddings. A senior engineer pushed back, commenting, "You're optimizing for semantic understanding but ignoring high-throughput latency constraints. You're attacking a Straw Man of 2010-era Collaborative Filtering; modern factorization machines are still the gold standard for the candidate generation layer." They were right. I rewrote the piece to advocate for a hybrid retrieval strategy _2. Learning to listen to this feedback specifically the "Steel Man" version of opposing views _3 is a skill that is universally applicable.
3. It is Version Control for Your Mind
Your past thinking becomes archived and searchable. This is more valuable than you think.
The Cringe Factor: Reading your thoughts from three years ago is like looking at a photo of your high school haircut: slightly embarrassing, but proof you’ve evolved. You can’t see how much you've grown until you have a record of how clueless you used to be.
4. The Network Effect
I am amazed at how many of my favorite people to chat with I met through writing online. The current of connection flows upstream, too—if you read something that resonates, send a note.
5. Managing the "Dormant Drafts"
The second most common question I get is, "How do you motivate yourself?" The answer is that I often don't. There are so many drafts that live, dormant in my draft folder. But eventually, something clicks, and I finish a piece in an hour. These rapid-fire pieces keep the momentum up.
How to Start (Without Overthinking)
It is hard to form new habits, but writing is the best investment you can make today. Here are a few heuristics for getting started:
The "Forklift" Rule
~ You must do the writing, not AI ~
Writing is exercise. If I brought a forklift to the gym and used it to lift weights, what would be the point?
Do not let AI generate your core text. When you offload the writing to an LLM, you are offloading the thinking, leading to cognitive atrophy. _4
However, AI is a wonderful editor. Once you have sweated through the draft, use AI as a linter. Paste in your draft and prompt it: "Identify places where I am not being clear or where my argument is weak."
Stop Worrying About Quality
Most writing isn’t great! If my hit rate is 1 out of 5, I’m thrilled. Some people worry about the risk of bad writing, fearing public embarrassment.
The Reality of Bad Writing: The risks are actually quite low. Bad writing isn't a car crash that everyone stares at; it’s a tree falling in an empty forest. It makes no sound, disturbs no one, and simply dissolves into the digital soil unnoticed.
Don't Over-Engineer the Stack
Don't overthink where to publish. Notion, GitHub Pages, or Substack are fine.
Essential: An easy contact form (email > comments).
Non-essential: Custom CSS, complex hosting, or RSS-only feeds.
Critical: Low friction. If it takes too many steps to deploy a post, you won’t do it.
I hope you start a blog this year. Or revive an old one.
Footnotes
On Cognitive Resistance: This is supported by the Generation Effect, a cognitive psychology phenomenon demonstrating that information is better retained when it is actively generated from one's own mind rather than passively read. See: Slamecka, N. J., & Graf, P. (1978). The generation effect: Delineation of a phenomenon. Journal of Experimental Psychology.↩
The RecSys Trade-off: For the engineers reading: The debate between Matrix Factorization and Deep Learning/LLMs usually boils down to the latency-accuracy trade-off in the candidate generation phase vs. the ranking phase. While LLMs offer superior semantic understanding, standard two-tower architectures (like Covington et al., 2016) often remain the baseline for the retrieval layer due to sub-10ms inference requirements.↩
On "Steel Manning": The intellectual opposite of the "Straw Man" fallacy. To Steel Man an argument is to address the strongest, most charitable version of your opponent's position. If you can defeat the Steel Man, your argument is robust.↩
On "The Forklift": This relates to the "Illusion of Competence." When an LLM generates a coherent paragraph for you, it feels like you did the thinking because you are reading it. In reality, you skipped the neural pathway formation required to organize those thoughts yourself.↩