Hey HN,
Infinite Swap is a game inspired by Kyle MacDonald’s One Red Paperclip and other bigger-and-better swap stories - you start with a green bottle cap and trade it with other characters in various locations all the way up to a house and beyond.
A while ago I got inspired by viral games Infinite Craft and What Beats Rock? and wanted to make my own ‘exploring a LLM’s latent space’ style challenge. After much experimentation this became a trading style game, where the LLM comes up with all the items, trading scenarios and value estimates, which ended up being surprisingly fun and somewhat of a roller coaster to play.
I found it was the most intriguing if I separated the item offer from its valuation - so I implemented a pricing oracle called “Sal” who gives this info with his, sometimes snarky, opinion after each trade.
The item offer mechanism goes deeper too - each character’s profile is combined with a set of hidden motivations. I then use embeddings and cosine similarity to represent the desire the character has for the player's item, and therefore what they are willing to offer in exchange.
Links to play: iOS (small number of ads, polished): https://apps.apple.com/us/app/infinite-swap/id6761694545 Android (small number of ads, polished): https://play.google.com/store/apps/details?id=com.squizsoftw... Web (no ads, a touch rougher than the mobile versions but still very functional): https://infiniteswap.app
Stack: GPT-4.1-mini for item generation, SDXL Lightning for item images, Kokoro for voices (the latter two running on my 4090 at home). Postgres + FastAPI on k3s for the backend, native Swift/Kotlin for mobile, React for web.
Keen for any feedback, bug reports, improvement suggestions, etc. I’ll be around for a few hours if you have any questions.
Daniel
I played it. It was fun. I especially liked how the price was assessed immediately after the exchange, which made it easy to understand.
Is LLM generated every time? If so, I'm worried that the API usage fees must be quite high.
Thanks for playing, glad you enjoyed it!
Re the API costs — this is a great property of the "explore an LLM's latent space" game. Once a character has proposed a trade for a given (player item x hidden motivation) pair, I cache it for use again in future plays. This means the inference cost per player trade round will drop over time. Costs are also offset by ads on the mobile versions (I plan to add them to the web version eventually).
Also, GPT-4.1-mini is surprisingly cheap for this operation as well - since posting here a few hours ago total inference spend has been in the low single digit dollars - so not really a concern at this scale.