I’m not that addicted. Am I? Link to heading
I don’t use agents for coding. I love my craft far too much. Don’t get me wrong, I’ve of course tried it. But every time I felt the worse for it afterwards, learned less than nothing, and ended up with code I neither owned or understood completely. Yuck. Nevertheless, LLMs have been sneaking into various workflows without my being completely intentional about it. My mantra has been to use it like Python scripts, “to automate (or simplify/streamline) the boring parts” while keeping it at arms length from everything I truly care about. But “boring” is a woolly and dynamic term, and I find myself growing lazy in more ways than I’d planned, as well as more easily bored. I suspect my brain is shutting down various faculties whenever I don’t bother finishing a sentence in my notes, summarizing longer parts of text or wraiting slopppy because some LLM will proof read things afterwards anyway. It’s a bit like when Frodo slips on his ring to get out of a tough spot. It seems to solve an immediate problem quite efficiently, but every time Sauron closes in (and it’s actually never worth it).
Enough of that.
Detox time Link to heading
Fasting is a quite efficient means to rid ourselves of various addictions, to intentionally starve our “lesser hungers” in order to bring our deeper ones into clearer focus. See, that’s not a bad sentence, right there, and no LLM refined it or articulated it on demand. I suspect this fast will do me good, and possibly unearth an uncomfortable reality that shows that I’ve become far more entangled in the quickfix life of LLM answers than I ever set out to be. I’m reminded of the many heartbreaking moments of both Blade Runner movies, when someone (something?) who thought they were Human found out to their own dismay that they weren’t after all, they were sophisticated machines programmed to act and feel like they were. But I am. Human, I mean. And I rather be that, it’s actually quite a privilege. And I rather be bored at times, than turn off more and more parts of my human brain to save… what, exactly?
So for the rest of November, I’ll do even the most mundane of tasks by hand. No GPT, no Claude, no nothing. Like typing out the front-matter (metadata) for this blog post:
+++
draft = true
date = 2025-11-09
title = "I'm taking a three-week LLM fast!"
description = "Getting my brain back well before Christmas 😌"
tags = ["passion", "AI", "lifestyle", "craft"]
+++
(I’ll change the draft bool before you read this, obviously, but you get the idea.)
A bit boring, yes, but I think it’s good for me.
(Btw, to clarify, I’m keeping my language servers running, and I’m still using electricity to cook food – I’m strictly talking LLM usage here, in case I wasn’t clear. How do I know if I’m getting my point across if I’m not asking Claude? I guess I need practice.)
And my book? Link to heading
Listen, I’ll probably run things by one of the usual suspects after a while, when this fast is over. I’ll probably ask my usual questions before publishing a new chapter. “Are all my hyperlinks working?”, “Did I sound over-confident in that last section?”, “Did I go too far making fun of that React framework?”. After all, proof reading and language in general is something an AI can actually excel at (second L in LLM, remember?)! But my favorite part of writing my Elm for React developers book is when actually when I’m putting on my usual playlist, and I’m all alone with vim, riding my commute train to work. Offline, being creative (I find those two are often deeply connected). More of that in these weeks, and hopefully after that as well. And if I feel stuck on a ride, completely at a loss as to how to continue, or which topics to cover first in the upcoming chapter? So be it. I’ll savor the feeling of mystery, enjoying or at least accepting that I have limits and I (and my projects) need time.
So that’s what I’ll do these weeks. Manual proof reading, manual checking for inconsistencies in my code samples etc. Gettng stuck, getting unstuck – the works. I need to work out the mental muscles made for carrying unanswered questions, because they’ve gotten terribly out of shape lately.
The end goal? Link to heading
To quote Bilbo, before he leaves the shire (and more importantly, his ring) behind:
I feel I need a holiday, a very long holiday, as I have told you before. Probably a permanent holiday: I don’t expect I shall return. in fact, I don’t mean to.
This fast is an end in and by itself, but I except to make some changes when it’s over as well. Perhaps I’ll time box AI usage more, like I do email? Email on Mondays, LLM tasks (gathered up in some batch throughout the week?) after lunch on Thursdays? Who knows.
I’m reminded of this concept I heard Cal Newport mention in his Podcast: “Hire yourself as your own personal assistant”. I’ve tried to do that in general; to shove all less-than-creative tasks into a separate list rather than be distracted by them throughout my workday. The idea is, of course, to still do them yourself, but to do them intentionally in a scoped batch once a week (or something). Then you a) get to do “deep work” more often, without expensive context switching all the time, and b) it becomes quite transparent how much time you actually spend on things that are less important than what you’re actually hired to do. Perhaps I’ll do something similar with my LLMs? Keep a list of things I consider asking them to do, but only actually interact with them in a given time slot?
We’ll see. I’m on day one still, what do I know?
But it feels good! Day 1/21.