Stream-of-Thought, with Stakes

I’m trying something different today. I’m using this web app to write this, set the timer to fifteen minutes. If I stop writing for more than 10 seconds, it deletes everything I’ve written.

This is a version of an older app called The Most Dangerous Writing App I think, something like that. This version was made by Dynomight. They have a good blog / newsletter.

The job search continues. I just applied for a role at Spotify today. I have to never put too much hope into any single job application.

This app is a little stressful. Can’t stop typing for 10 seconds.

I’ve increase the amount of time I committed to working on my job search from 30 minutes a day to one hour today, committed via Beeminder of course.

I’ve started using tmux for spinning up development environments in my terminal. I switched from warp.dev to ghostty. I run `dev`, a custom shell script in my bin, and it runs a whatever-project.sh script that loads the tmux session. The left column is always Claude Code, run with --continue and falling back to straight claude. In the right column I’ll have one pane running any dev servers, one pane running tests, and one just a straight terminal. I’ve also played around with having the middle column be hunk in watch mode so I can see the changes on the branch as I work. Not totally happy with how that works yet though so only have that in one of my environments.

I’ve expanded my job search to include not just jobs that would relocate me to Sweden but also jobs that are fully remote. That way I can still have the freedom to travel even if I don’t get a job in Sweden specifically.

With projects like this where I really want to keep velocity I find that it works quite well to have multiple Beeminder goals supporting the project from different angles--in this case, daily time, action points, and applications submitted. When one goal starts slipping in its effectiveness, the others are a backstop while I diagnose and tweak. And it feels good to get ahead in one goal even if it’s another goal that’s the reason I’m getting ahead.

I’ve experimented a bit with working on a treadmill again. A few years ago I had a treadmill that fit under my desk, but I found that keeping it working was difficult. Now I have access to a full-size treadmill where I’m staying. I discovered that there are universal attachments you can buy for full-size treadmills that add a desk or a laptop stand to them. For an experiment I tried just using a board, and it did seem to work ok, especially since my programming work tends to be give commands, monitor, give more commands, etc, so I don’t need to be hands-on-keyboard constantly. I don’t think I’ll be doing it much, though, since I’m generally only spending about an hour on the treadmill over a day.

My M5 MacBook Air is proving to be quite functional as a work laptop. This is the first Apple Silicon computer I’ve owned. I have the 32gb model. I haven’t experienced any performance issues, and have even used some local Ollama models for certain things. The only real issue I’ve had is Firefox filling all my memory after days or weeks of not restarting the browser.

I do feel like this experiment of using this writing app for writing the newsletter is working fairly well. It’s basically what I used to do when I would write daily pages—just write, write, write, no stopping, even if I’m writing about how I don’t know what to write about.

I do still feel that editing is where I currently fall down in my writing. That’s where the anxiety can start becoming a problem.

Maybe writing by hand is an even more important practice than it used to be, now that pretty much anything can be theoretically done with AI. A way to make sure we don’t forget how to think.

I know some people have advocated for also dedicating some time to hand-coding so you don’t forget how to code. I’m not sure how I feel about that. Is that like advocating for spending time each day writing by hand or traveling by horse? Like, yeah, that’s a skill you could lose, but it’s so superseded it isn’t actually a practical problem if you lose it?

I guess the question would be what are the stakes if I’m wrong. Like if I forget how to code by hand, completely lose that skill. How would that hurt me? How would that hurt my work? Would I be less good at determining the quality an LLM outputs if I’ve forgotten how to code by hand?