A lot happened in the last month.
I got a cold. I got the flu. I saw my family. I got another cold very briefly. And I started a new job.
I also spent hours going in circles on Chapter 2.
What’s rough about Chapter 2 is that it’s chock full of small examples. Each one has to be vetted, linked up with the narrative, and rechecked to make sure it builds towards the broader points I’m trying to make. You can get a small taste of this in the first draft of the chapter.
It feels like I’m making progress. A month ago I was still making large, structural edits to the chapter. Now I’m doing a lot more polish. I probably spend too much time trying to find the right words, but when I feel like the words are wrong, it is, for me, a useful signal that the content is still not quite right. So I write and rewrite over and over.
Claude helps a lot, although I find I’m still finding the best way to work with him. He will sometimes indulge my ideas too much and help me go rapidly down a path that’s going to fail, but at least we figure it out together more quickly than I would on my own.
AI
Speaking of Claude, he continues to integrate himself into my life more and more every day. I’m writing a lot of code in my new job, so I have a lot of chances to interact with Claude and other AI tools.
I know I wrote about using Claude last time. But he’s become an important part of my life that, and I’m still figuring out how best to use him, so I’m going to do it again!
The way I think of it, Claude is my work buddy. I can do with Claude what I find almost impossible with humans: pair programming, or at least the version of it he’s capable of. I ask him to implement something, he comes back with some code, I bug fix and edit and ask for corrections, he obliges. Sometimes I write code and it doesn’t work, and he looks at it and finds the bugs or suggests improvements. It’s a great relationship that’s making me both more productive and a better programmer, which is somehow possible after more than 30 years of programming!
I’m less a fan of other AI tools. I’ve tried all kinds of AI integrations in my text editor, and they are fine, but none of them beat chatting with Claude directly. Maybe it’s just me, but the cost of copy/paste for me is low relative to the benefit I get from the space between Claude and the editor that all the integrated tools lack. So for now the only thing I’m using in editor is Copilot code completions, which are right about 60% of the time, and close enough another 20%.
Other writing
Last month I wrote a short piece about management philosophy and posted it on Substack. You might note that the newsletter itself is hosted on Substack, so you’re probably surprised to know I did this. That’s because when I posted it I turned off the feature to email and notify you all about it.
I instead shared the post on Linkedin.
I did this because my original premise for using Substack was to use it as a monthly newsletter. My plan was just to send you all 1 email a month max.
But, here’s the thing: I’m going to write other stuff. I could post it elsewhere, but then that decreases the opportunity for folks to join the newsletter and ultimately learn about my book and, most importantly, the ideas in it.
So, a question for you all: How would you feel if you got 2 messages a month from me? One would continue to be this monthly newsletter, and the other would be a short article about something interesting, like revisiting ideas from some of my most popular Less Wrong posts and new ideas I want to work out in 1-2k words.
Let me know your thoughts in the comments. If you’re reading this, you’re probably one of my current 12 subscribers. All of you are precious to me. You’re the seed of what could one day be a vibrant community. But you’re a seed that will sprout only if I tend to your needs. So tell me what you need!
I just subscribed today, but two messages per month sounds perfect to me.