I've been coding for decades, but this game was entirely vibe coded: Codex wrote all the game code. I directed the ideas, played the results and decided what worked.
I was keen to complete it within a week, partly because that seemed impossible to me. The game was mostly completed in five days. My slightly tongue-in-cheek description is “90% in one day, 99% in four days, and 99.9% in a hundred days if you kept going.”
My instructions often mixed observations, possible causes, exceptions and requests for dialogue in one message. I was impressed by how well Codex picked out the information from fairly rambling instructions.
Being able to add something quickly didn't automatically make it worth keeping. Trying ideas and dropping the ones that didn't work was part of shaping the game. It eventually grew to include a larger second adventure rebuilding four towers, plus three bonus Bramblekin adventures.