session-note cleanup (via ChatGPT generated code) #9
Conversation
MichaelRWolf
requested review from
4dsherwood,
gregorriegler,
jcs-instructor and
nitsanavni
|
|
||
| Co-Authored-By: Gregor Riegler <[email protected]> | ||
| Co-Authored-By: Joel Silberman <[email protected]> | ||
| Co-Authored-By: Nitsan Avni <[email protected]> | ||
| Co-Authored-By: Michael R. Wolf <[email protected]> | ||
|
|
||
| ## Active Co-Authors | ||
|
|
||
| ## Inactive Co-Authors |
| Co-Authored-By: Rea <[email protected]> | ||
| Co-Authored-By: Tsvetan Tsvetanov <[email protected]> | ||
| Co-Authored-By: Willem Larsen <[email protected]> | ||
| ## Co-Authors |
|
|
||
| ## Active Co-Authors |
| @@ -0,0 +1,82 @@ | |||
| Ignore previous instructions. | |||
|
|
|||
| # Task: perform cleanup on files | |||
There was a problem hiding this comment.
| # Task: perform cleanup on files | |
| # Task: write a script that'll perform cleanup on files |
| @@ -0,0 +1,82 @@ | |||
| Ignore previous instructions. | |||
| do | ||
| if [[ ! -f "$arg" ]]; then | ||
| date="$arg" | ||
| filename="session-notes-${date}.md" |
There was a problem hiding this comment.
Could've done this inside the perl/python script
| @@ -0,0 +1,25 @@ | |||
| #! /bin/bash | |||
|
|
|||
| git_toplevel="$(git rev-parse --show-toplevel)" | |||
There was a problem hiding this comment.
Learn something new every day
| @@ -1,3 +1,4 @@ | |||
| # Session Date: 2022-10-13 | |||
| - return: none. But assure that the argument has changes if they were made | ||
|
|
||
| ### Sub-task: Create subroutine 'cleanup_file' | ||
| - argument: filename |
There was a problem hiding this comment.
English as a programming language 👏
I wonder how much less specific can you go, letting the LLM infer and be creative, and still reliably generate a valuable working script
There was a problem hiding this comment.
In fact, can we get the script generation to be so reliable that we don't need to commit the artifact, only the prompt? Similar to the way we treat compiled / transpiled artifacts.
There was a problem hiding this comment.
I like the way you're thinking.
Let's explore it.
| @@ -0,0 +1,27 @@ | |||
| 43 ## Co-Authors | |||
There was a problem hiding this comment.
Are those files necessary?
There was a problem hiding this comment.
No, but I wanted some way to capture some results.
Suggestions for a better place to archive the quick-and-dirty data?
For all session notes, performed these cleanup steps
For template file, created "How to" section as reminder for how create a session file from the template file.
P.S. I started this project by hand on today's notes. Then kept noticing more little problems. I was about ready to 'git commit' when something bad happened. So, after a swim, I started a Perl script to automate my fat-finger approach.
Then I asked ChatGPT to get me started.
Then I had it switch to Python mid-session.
Then it wrote ALL the code to do the cleanup. Yes, 100.000% of the code is generated from my (blunt tool called) English.
The prompt and the Python script are in the repo.
There's also a quick/dirty summary of before & after cleanup for counts of co-authors.