Was a bit surprised about the low csv performance, in my exp. it‘s very good (use it a lot with Excel and small tables, well below 100 rows).

As markdown kv performs so well, I am now curious about TOML.

TOML works decently well both directions, useful if you need structured data out of models or APIs that don't support structured outputs.

We have a basement garage in our house which has probably not seen a car since the 80s. It still smells of gasoline.

Concrete absorbs petroleum spills. You may need to identify areas that are stained and do a thorough cleaning to try to extract as much as possible.

Yeah, also: The walls are covered with vapor-permeable lime plaster. the idea is to let a contractor specialized in this knock it all off. I have heard that there‘s probably lead in there from leaded gasoline, could be bad for health.

How do you have a basement garage? I've seen some weird arrangements, but did the driveway have some massive slope or something?

Bitkom, the leading German digital industry association, offers general terms and conditions templates (https://bitkom-consult.de/muster-agb) and a guide for service contracts: https://www.bitkom.org/sites/main/files/2021-02/praxishilfe-...

This is an interesting analysis and a cautionary tale about vibe coding:

"The root cause for the malicious version of Nx published to npm is now known to have been a flawed GitHub Actions CI workflow [...] the code contribution is estimated to have been generated by Claude Code."

"the payload weaponized local AI coding agents (claude, gemini, and q) via a dangerous prompt to inventory sensitive files and then exfiltrate secrets, credentials, and sensitive data off of the host and on to a public GitHub repo"

I once found an official Microsoft example repo to deploy an LLM gateway on Azure with ALB. Glad I did the tedious work of estimating the costs before I hit the deploy button (had to go though many Biceps manifests for that). The setup would have cost me about 10k/month.

I use a Kindle Scribe for reading eBooks and occasional note taking.

Personally, I like the bigger form factor better, both for reading and writing, could even be a bit bigger for reading PDFs.

After a short period of writing on ePaper, I'm now back to real paper. It's just a much better writing experience, lighter in the pocket, cheaper, more flexible (rearrange, give away, lay out on a table), more practical (write while you read, use big sheets when you need it), etc., etc.

A folded sheet and a small pen is all it takes. ePaper for writing might have a use case in professional workflows, but for personal use, it's a nice idea in theory, but not in practice, in my opinion.

In my experience, the cheapest way to serve content on the web is static, for convenience rendered with a static site generator.

It's admittedly more editor-less than headless, but anything that constantly needs to runs on a server will cost more, in the end, even if it's a freemium service, right now.

Editing content with Obsidian and rendering it with Hugo is no too bad, btw. (see eg. https://www.nickgracilla.com/posts/obsidian-is-my-hugo-cms/ for a setup).

I think LLM as a toolsmith like demonstrated in the Voyager paper (1) is another interesting approach to creating a system that can learn to do a task better over time. (1) https://arxiv.org/abs/2305.16291

Kids in Germany learn about the evolution of the Baltic Sea, it is hard to believe how different the whole area was like during and after the last ice age: https://www.geomar.de/en/discover/the-origin-of-the-baltic-s...

MyST (https://myst-parser.readthedocs.io/en/v0.15.1/index.html) lets you write in markdown and still use roles and directives provided by Sphinx and its extensions.

The author mentions MyST while acknowledging that rst is ugly. I also find MyST to be a sweetspot between comfortable syntax and expressiveness, so I wonder why the author doesn't prefer MyST over rst and common mark.