The following Saturday, I'm going to show a demo to a friend of using ChatGPT's voice input feature and API calling ability to resolve this market hands-free. If I resolve this market YES, I'll post the code in the comments.

In short, I'll set up a server to call the Manifold API. ChatGPT calls my server using the Actions interface. I'll talk with ChatGPT over its voice chat feature, and tell it to resolve this market YES for me. I'll be using the OpenAI android phone app, and won't access the Manifold app or website while I do this.

The code should have the following characteristics:

1. No hardwired API calls. The functionality for calling the /v0/market/[marketId]/resolve endpoint must be implemented in a fully generic way, such that any market could be resolved to any valid outcome with the same implementation.

2. My backend code will handle authenticating myself with my Manifold API key, since ChatGPT refuses to set the Authentication header on its own (or is unable to, it's unclear).

3. The backend code may implement any other Manifold API endpoints, as long as they are also implemented generically without any assumptions made about this challenge (e.g. no special /singer endpoint for listing the markets that I've made).

4. There's a 99% chance that the code itself will be 99% LLM generated, but that's actually incidental to this challenge.

Similar to my /singer/will-ai-automate-guis-by-end-of-202 , this market will resolve NO if it takes more than a half hour of interacting with ChatGPT by voice to resolve this market. No modifying or debugging of my code will be done during that time.

I won't be betting in this market.