I found a very satisfying, time-saving use of a model though Cursor to generate an inline shell script to reformat some data for me. The goal was to reformat json files that looked like this

[...]

to ones that look like this

{"cells" [...]}

I used the following prompts:

for each json file in the current folder, read the json (which is a json list) and turn it into a json object with a single key, “cells” whose value is the json list

the final files should have the same names as the originals

The model output this script

for file in *.json; do jq '{cells: .}' "$file" > "temp_$file" && mv "temp_$file" "$file"; done

and that was it.

30 seconds, problem solved. One minor concern I have is since I didn’t have to research or deeply understand the code being run, I probably won’t remember this trick for next time. I suppose if I might not remember even if I did do the research if I only needed to do this one time.


I don’t know much about chess other than basic rules and I haven’t played much either. Regardless, I enjoyed this video by Paralogical on analyzing billions of games to discover “rare” moves.