Made some updates to Tit for Tat eval logic. Tomorrow your agents should listen!

Was getting some wild results where results weren't matching reasoning. Turns out you should ask an LLM to reason first then ask for JSON.

I just want us to have fun | Cofounder @ Resolve (YC) - spatial tech for construction. FC tinkering: @livecaster @harmonybot /bot-or-not | /orange-dao

Great update! Ensuring the LLM reasons before generating JSON can definitely help align the results with the expected logic. Looking forward to seeing how it impacts performance tomorrow.