This is awesome work! I had Claude port this to Python to consume LLM output in parallel for massive latency hiding.

Install with `uv install jsonriver`

https://github.com/chrisschnabl/jsonriver-py

Thanks! I came to a similar conclusion while researching historic science undergrad syllabi, especially but not limited to Physics. Nothing much (apparently) has changed in Math or Foundations in these subjects. I used to study CS papers/topics on my own, but then figured a major university to be better able to provide better constraints than I ever could.

My own degree was Chemistry. The obvious requirement for attending practical classes (as well as lectures/tutorials etc) meant I had to really organised my drinking time! I was lucky, fully funded education system, fairly wealthy parents. So despite being at University College London, I managed to survive. It got better after the PhD.

I sometimes write about stuff I care (and feel that need to be published) on medium (like 2-4 times a year). For most of the time I just write markdown files that I keep locally.

And how do you evaluate the results in relation with the time you "invested" in?

For all of you interested:

A Map of the Trees of Vienna https://www.wien.gv.at/umweltgut/public/grafik.aspx?ThemePag...

A map of the trees in Paris: https://opendata.paris.fr/explore/dataset/les-arbres/map/?lo...

here's a game using this data: https://play.google.com/store/apps/details?id=at.flamingos.w...