alamb commented on issue #24569:
URL: https://github.com/apache/datafusion/issues/24569#issuecomment-5401164764

   
   > One thing I've been thinking about is how we should write tutorials 
differently in the age of LLMs. AI can explain things and often give shockingly 
good explanations, so perhaps tutorials should become more concise and focus 
more on intuition and the key ideas, rather than trying to be complete. 🤔
   
   Yes -- I have been dreaming about this too and I have a bunch of ideas that 
I want to try and help out with. I think this datafusion-cookbook should be 
primarily targeted at people building software with LLMs / coding tools
   
   For example, I am imagining a user providing a prompt like this
   
   > "make a duckdb style CLI with variant support for semi-structured types 
and auto complete SQL optimized for reading from object store"
   
    I tried something like this at a seminar I was at recently. While claude 
was able to implement it, I found it took a while and took a bunch of iteration 
and instruction. I am feeling if it was complicated for me to do it, the 
barrier is still way too high
   
   So what I imagine is "testing" this repo regularly with some pre-cooked 
prompts and then figuring out what is missing to make the process faster and 
more efficient. 
   
   For example, one thing that is complicated, is figuring out versions and how 
to integrate them. One  thing I could see is a table / listing of all the 
various datafusion "plugins" that exist (e.g. datafusion distributed, flight, 
etc) as well as pointers to version compatibility and installation instructions 
/ examples
   
   
   
   
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to