Happy new year! Hope you spent the magical time between Christmas and New Year’s doing whatever you want — that’s what it’s for.
What I did? I upgraded my load-and-lights-bearing Raspberry Pi 4 from a 32-bit OS to 64-bit. That went ~relatively smoothly: I used Claude to write me up a checklist of commands to run and walk me through them one by one. “Do you now want to proceed with restoring the Docker volumes?” Only problem was that Home Assistant cannot restore from backup if it runs in Docker. Uh-oh. Still, Home Assistant is one of the best open-source projects I’ve ever used.
I also spent a day and $3 using aider with Claude to build a multi-player “emoji garden” in Rust without writing a single line of code. If you haven’t, try aider. Put Claude’s API key in, set yourself a timer for 45min and build something new. You might experience some magical moments that, I think, contain a reflection of the future.
Talking about Home Assistant: voice assistants in Home Assistants are apparently a thing now, wow.
What a url: sacred.computer. It’s “an open-source React component and style repository that helps you build […] with terminal aesthetics.” Looks nice.
This Stanford lecture on building LLMs was incredibly good. Nitty-gritty at just the right level of abstraction. Lots of “ahh, so that’s how they…” moments. I want to write a separate thing about this lecture because it’s so good.
Spent some time in the past two weeks digging into transformers and how attention works. This video made think: what a miracle!
Pay attention to your thoughts immediately after reading the next sentence: “This is just a hobby of mine, that I thought might be interesting to a lot of people.” Now scroll through the hobby website from which I quoted it. Isn’t the Internet a marvellous place?
This post by Anthropic on how to build effective agents has been shared quite a bit in the last two weeks but I need to re-share it here again, because it is remarkable for a post like this to be so free of bullshit, leading with what is not an agent, recommending when to not use agents, and to deliver some really good insights.
What’s magic? “Sometimes magic is just someone spending more time on something than anyone else might reasonably expect.” Great little post.
This is very, very good: “If you work on anything worthwhile, sooner or later people will care about it and will want you to send progress updates. These could be quarterly investor updates, weekly updates to your boss, emails to adjacent teams, etc. Here are tips on how to do this well.”
Good questions to ask: “If generating code is no longer a limiting factor, what types of programming languages should we build?”
Vicki Boykis before listing her favorite books of 2024: “I also need to read fiction because if I only read tech books, they take away my creative energy rather than grant it, as fiction does.” Nodded so hard while reading that my headphones fell off (not really) but then realized that I don’t read that much fiction anymore. I do read a lot of non-fiction-&-non-tech, though, and I do know that reading David Foster Wallace on Federer gives me something I can’t find anywhere else.
One of my favorite 2024-in-review posts is this one by Nicolay Gerold. Humility breeds wisdom, is what it tells me.
More wisdom in this wonderful piece by Armin: “Life isn't about sampling everything; it’s about making deliberate choices and committing to the ones that matter” and “There is a narrative that working hard is inherently bad for your health or that long hours lead to burnout. I disagree.” and “If you end up doing things you do not believe in, it will get to you.” and… I will stop now. Read it.
Michael Lynch’s Rules for Writing Software Tutorials is a goldmine. I agree with all of it, except maybe the bit about how to format shell commands, since I don’t like copy & pasting ten lines of &&-joined shell commands, but instead want to execute them one by one.
Beautiful piece by Maggie Appleton on being pregnant: Growing a Human: The First 30 Weeks. The sections on “Natural” and “Traditional” brought up a lot of memories from four and eight years ago when my wife was pregant. I had some (to my mind) under-appreciated rants on “natural” back then, including little detours into “full of chemicals.” Now I’m waiting for Maggie to write about doulas and what the first six months are like with the baby. Seldom does the rubber hit the road as hard as when it’s 3am, you’re holding a crying baby in your arm, it’s so god damn hot in this room oh my god, you and your partner are getting angrier and more frustrated by the minute because you both don’t know how to make the crying stop, and the "this is what I'll be like as a parent" rubber hits the road of "this is what I'm actually like". Anyway—
pISSStream is “a menu bar app that shows how full the International Space Station's urine tank is in real time.” Now you know.
Here’s an intro for you: “In this tutorial, I will teach you how to write your own virtual machine (VM) that can run assembly language programs […] The final code is about 250 lines of C" — Beautiful, little program. Love the formatting of the page and that it’s literate programming.
Love this list and all that are like it, even though I might not agree with every single point. (Imagine a CEO writing a list like this about company culture!)
And here’s a list we should probably all keep: “In 2016, I worked with a really sweet guy who kept a running list in his Notes app simply titled ‘Things I Like.’ I copied him that year and have kept it running ever since.” (I keep a list of good names.)
Jon Gjengset on “what’s the right number of hours to work in a week” and how the answer to that question “varies not just from person to person, but from week to week, and even day to day”, along with his thoughts on it. And… I don’t know. I don’t know. I think having a number like that — 40 — biases and distorts everything. Not saying we should, but what if we remove the number?
Very often in my life I wondered whether I shouldn’t get into sports. Just seems like a fun thing to follow your team throughout the year and always have something to talk about with others. Then, while reading this and hearing it in my head with the voice of a sports commentator (listen: “He generally tries to keep a symbol <> file mapping in his head. As long as whatever he's working on is loaded in his mental RAM, he manages fine-ish. But when not, after fumbling around, he has to bring up an external grep tool.”) I realized I might have found my sport already. Then I scrolled through this list of most active GitHub users, recognizing many names, nodding along, thinking “ah, didn’t they switch over to…” and knew: yes, this is my sport.
A title that makes you click: I am rich and have no idea what to do with my life. But there’s more to the post. For example, this paragraph about culture: “I learned about the power of urgency and having an undeniable mission. Not by reading it somewhere. By experiencing it. I came to realize how laughable my robotics stint had been in comparison.”
Stuck with me: “What would it mean to be done for the day?” (Reminds me of Jason Fried once saying that he generally aims to do a “good day’s work”.)
Succinctly put: burn the playbooks. “Someone does something for love of the game or out of pure curiosity. […] Others see the success, deconstruct how they did it. Others still repeat those same things, joylessly, without the original spark. Create a cheap shadow of the original.”
Another sign of the times. Robert Ghrist wrote a book on Linear Algebra and put this in the introduction: “The writing was assisted by Claude 3.5 sonnet, trained on my previous books for style. [...] Gemini Experimental 1206 was an especially good proofreader and converter to markdown. Exercises were generated with the help of Claude and may have errors. This project was begun on November 4, 2024. The first edition was submitted to Amazon Publishing on December 28, 2024. Fifty-five days: impossible without the creative labors of Claude, to whom the author is most grateful.”
