This was my first attempt at using Google Gemini CLI to one-shot a full working solution.
tl;dr: I wrote the specifications for the program I wanted (A Server Side RSS Reader similar to this). The entire spec I wrote can be read in GEMINI.md.
Then I executed Gemini code with the following command:
gemini -f GEMINI.md
I said to various requests like writing to the file system, executing generated code etc.
Gemini ran the code it generated, and when it failed, it went back to editing the code, and reran the script.
After a couple of programming bugs (handling null
attributes), and adding library deendencies (lxml
, feedarser
), the program was done.
I captured a couple in-process screenshots of the program in action.
I checked in all the generated code and the output of the program in this commit