I wanna share some useful Cursor insights with you.
I was having error after error trying to create a new field under my users table (postgres). After 20 minutes and a half dozen failed attempts, I checked the docs Postgres | Documentation | Payload and one prompt later, it all worked. New field was added.
So I asked Cursor in the console:
so lets review the errors we encountered and the solution.
why were we having so many failed attempts?
what did you learn that you can share with me, so i can better prompt you in the future?
It then proceeded to breakdown our journey and the lessons learned.
So I linked it to .cursorrules and told it to modify it with what weāve learned.
Never tried that before. Pretty freaking cool. Iām loving Cursor more and more everyday!
@jjaimealeman Hi Jaime, thanks for this. Would you be able to share you .cursorrules file? It is always nice to see how other people are doing things. If there is private info and you cannot share, I will also understand.
giving it the good olā ā and what did we learn, is there anything you would like to add to our rules to do better next time ā proved quite fruitful
Thanks for sharing this. Are you using a tool or a certain approach to generate this JSON? My rules file is just MD and Iām curious about the advantages of this structure.
Funnily enough, I had a āchatā with Cursor about which format it preferred, and it said the .json, while well-structured, was not as easy to follow as the markdown file. It said the markdown, being in a natural language, was easier to read and comprehend.
I copied my markdown formatted cursorrules into ChatGPT and asked it to format as a json structured object. Then named them .cursorrules1 and .cursorrules2. One was markdown, and the other json. It still preferred the markdown.
So thatās the one I kept and the format I continue to use.
Thereās a thread on here about the format, and it seems many users use json. So I really donāt know?!
Can you try and do the same as I did? Make one in markdown, and the exact same rules as a json file, and ask Cursor which it prefers and why?
Iād be interested to know what you and others find out.
However, I have been fiddling various .mf .rmd and json dialogues with all the GPTsā¦
Here is my off_the_cuff subjective anecdoteā¦
for data that is slurped into a project/workflow/context that is structured DATA - jsonā¦ for declarative, narrative, @docs, architecture, schematic ā¦ i,e, Knowledge Conveyance, Agentic Controls: .md (.rmd in my pref - as I like to tell it things like "include mermaids for all schema, logic, and convert to svg as well.
But like I showed, having it tell me stuff about the project that I can render to a synopsis HTML in Rstudio is a pretty dope workflow Ill be automating as a park of my workspace settings:
