LLMs playing around in the Glitch editor

last thread: A bot for Glitch's editor

without-with-first

Google Gemini 1.5 Flash taking a whack at a programming task on Glitch. Also pictured: me typing bad.

I had this code for programmatically watching Glitch’s OT server. since then, Google put out an LLM API with a free tier, so here it is playing around in Glitch

super janky, it sometimes likes to repeat the question, sometimes likes to put extra comment slashes, sometimes likes to start repeating the whole file :skull:, but sometimes when you’re recording, it cooperates.

code for this:

maybe it needs a more explicit prompt, the model is very used to replying in markdown. people say giving it examples work, but I have no idea how to do that when questions and answers are many lines long and have their own internal formatting. AI people please help :pray:

edit: Gemini API Pricing  |  Google AI for Developers  |  Google for Developers you need a google api token to set this up

5 Likes