Two GPUs, four new tools, and a 1959 paper

Hey friends,

This fortnight started with three days of Obsidian and Claude Code configuration and ended with a 1959 paper, making me feel slightly embarrassed by how impressed I was.

What I’ve Been Up To

I finally got the full PhD setup working the way I wanted it. Obsidian for the PhD vault and the content side, both connected to Claude Code and integrated with my simulation project repository. Three days to get it right. The point was to be able to go from research output to a public-facing blog post with far less friction, and that part works now.

I also worked out how to run parallel simulations on the two GPUs in my workstation via a short bash script. One run per GPU, both launched at once. Almost certainly something I should have done six months ago for quick parameter scans, before committing anything to the HPC clusters.

​Feedly is running with a few feeds to stop me from spending entire afternoons hunting for relevant papers and staying on top of the industry news.

Found NotebookLM as well, which I have mostly been using for my PhD proposal: uploading papers and chapters, generating flashcards, quizzing myself across a cluster of sources at once. The audio explainer feature turns a couple of papers into a conversation. Good for absorbing context while doing something else, like the commute to uni.

Also published a blog post on keeping plot styles consistent across scripts without repeating yourself, if that is your sort of thing: one shared theme file to rule them all.

Worth Reading

​A note on a method for generating points uniformly on n-dimensional spheres (Muller, 1959)

I needed this for generating particle orientation distributions in my research. The method fits into a few lines of code, works cleanly, and I was half-expecting it to be a difficult setup. If you ever need to sample uniformly from a high-dimensional sphere, this is the one. I will write it up properly with the code and a visualisation in a short post soon.

Quick Thought

The whole point of the Obsidian vaults and Claude Code setup is to make it less painful to share what I am actually working on with people who are not physicists. The infrastructure is in place now. Whether I follow through consistently is a separate question, but I have removed most of the excuses.

Since this is the first issue I have actually sent consistently, I am curious what you are here for. Takes one click.

Until next time, Julia

You’re getting this because you signed up at juliafrank.net. If this is not your thing anymore, you can unsubscribe below.