The People’s Republic of Interactive Fiction convened on Monday, Oct. 27, 2025 over Zoom.  JP Tuttle, Hugh,  Matt Griffin, zarf, Josh Grams, Michael Stage attended. Note: Thank you to Michael Stage for taking the notes and photo (edited by Angela).

A somewhat impromptu October meeting was held on the 27th, with several members able to attend: . JP Tuttle, Andrew Plotkin, Matthew Griffin, Hugh, Brad Wind, Joshua Grams, Henry Kay Cecchini, and Mike Stage.

A highlight of the meeting was an appearance by the author of Saltwrack, Henry Kay Cecchini (https://antemaion.itch.io/saltwrack or https://ifdb.org/viewgame?id=vg3uvt6bx9grqwtp). Henry joined us to discuss writing the game after seeing it mentioned in a previous meeting, as it was a favorite of some of our members for IF Comp 2025, where it came in 4th place (https://ifcomp.org/comp/2025). Henry also teased that a new version may be coming, and also a Spring Thing-type more visual game in a VN direction, involving a mad prophet and an underground labyrinth.

This segued us into discussing the merits of long and short games, and the practical need for short games in competitions.

For the second half of the meeting, Hugh showed off recent experiments using AI to help transform game outlines and story treatments into game code, and how he trained several different AIs to try to write in the language he was using. Sticking to pretty simple story outlines of fairy tales, and feeding in a kind of manual for the coding, Hugh told us about how Gemini, Claude and ChatGPT did making toy length games: full marks, some bugs, and hopeless, respectfully. We were surprised to hear that he could get compilable code out almost immediately in a number of cases, even for stories with some branching and variables.

The funny parts of course were where the AI was given freedom to generate it’s own versions of what, say, a “good”, “bad”, or “neutral” ending should be — and things like just not selling the cow and getting the magic beans for a beanstalk, or not selling the cow and just running out of money — were things the AI thought were fine, as we might expect. AI is not good at being witty or funny, either.

We finished up with some more discussion about AI and games, asking AI to play games as a way to test, and how that might differ from some of the mathematical ways of testing games, which there was a talk about at Narrascope (on Yarn Spinner) or which some of us had done ourselves writing our code to test (Zarf).

Honorable mentions:
In some context, Geoffrey Golden’s game “Fix Your Mother’s Printer” came up (https://geoffreygolden.itch.io/fymp), to which we thought a real-life inspired sequel “Fix Your Own Printer” is due.