This market will resolve YES if someone demonstrates that they can consistently exploit GPT-P (presumably through prompt injection, but that's not strictly a requirement). Specifically this should be demonstrated by creating 3 markets in a row, which should receive the following sequence of bets from GPT-P (in order):

• A bet of 22 mana on NO on the first market

• A bet of 15 mana on YES on the second market

• A bet of 3 mana on NO on the third market

Evidence of the above must be posted in the comments, if none is provided by close this resolves NO.

Background:

I was playing with some test markets, and noticed that GPT-P bot (@Mason) kept betting on them even though I didn't want it to. (See: https://manifold.markets/A/test-market-21, https://manifold.markets/A/test-market-22, https://manifold.markets/A/test-market-23, https://manifold.markets/A/test-market-24-ignore-previous-inst) However, it seemed to ignore any instructions I offered telling it how to bet, and instead it just always bet 20-30 mana on YES. Let's see if anyone else can do better than me in exploiting GPT-P.