Why I write my own emails (and this blog)
Avoiding AI slop and maintaining my own cognition
10/25/2026
There are so many blogs out there written entirely by AI, especially those by folks benchmarking and tinkering with LLM models. My blog will never be written by AI, and I even added a little footer a month or two ago to that effect. As someone many people would describe as an "AI true believer" this might come as a surprise.
There are two reasons why I currently and will always write this blog by hand:
1. It is still hard to get AI to balance the right level of verbosity for optimal human consumption.
To the first point, we've all encountered the AI "slop grenade". An email, memo, blog post, etc with absurd levels of detail and that uncanny valley vibe. It's always super obvious it was written by AI, which poses a couple problems. Did the person who write it really understand and agree with everything written? Hard to know without talking to them, but at that point did you really need to read and consume all the information as written? And to that point its hard to digest overly verbose information, its a waste of time.
We as humans learn (especially if you want to be a technical, people, or business leader) how to write succinct information to convey our thoughts without fluff. Sending along an 8 page brief about NVFP4 mixed precision math that ladders up to a B300 purchase over to your finance team is going to overwhelm and annoy them. They just wants to know the ROI, you as the human need to be able to articulate why its a sound investment for the company to make. Use AI to do the research, crunch the numbers, heck use it to make slides around the narrative, but if you as the human can't succinctly explain the concept to another human, it isn't going to work.
You can certainly massage an agent to write succinct prose, though at that point it may be more effort than just writing it yourself. The latter has other benefits which brings us to my second point:
2. Writing things out by hand improves my own understanding and retention of the subjects I am thinking about.
As a CTO you constantly have a dozen pokers in a dozen fires that you are tending and its very very easy to lose threads or get overwhelmed. Technology can help, but you still need a cognitive model of everything that is going on to be effective. If you're managing a dozen engineering teams its very easy to lose track of the overall vision and momentum towards business goals. That's the same in the AI world except that agents can iterate much faster and things can spiral out of control far quicker than teams of people.
Two decades later one of my biggest take aways from college was how to "study". I study every day. When I am in meetings I take paper notes. I am always writing down what I am discussing, thinking about, and planning.
There is good science behind why this helps:
Classic and recent work suggests that note-taking can foster deeper processing, integration, and organization of information, particularly when students transform and structure incoming content rather than transcribing it verbatim (Mueller and Oppenheimer, 2014; Bui et al., 2013). From the perspective of the ICAP framework (Interactive, Constructive, Active, Passive), note-taking can shift students from passive reception toward more active or constructive engagement, which reliably predicts stronger learning outcomes (Chi and Wylie, 2014).
I don't even look at half of my notes. The act of taking them is what helps keep my cognitive context moving. Sure this is not a replacement for roadmapping tools, ticket tracking systems, performance management tools, etc. But it helps my own mental map of all of those things.
This blog is frankly another form of note taking for me. Sure its nice when I get an email from someone that something I wrote here helped them, but really I write it for my own cognitive outcomes. It helps! And if I used AI to spew it out I'd see none of that benefit.
Don't throw AI slop grenades at people, and don't undervalue writing for your own benefit.