Social agent debugging

[gabrielfior]

As raised by @kongzii , the social agent is tweeting odd things.

Below one can see the rationale (for think-thoroughly (TTA), agent that Social Agent acts as a PR representative for). Problem is that the probability of Trump being elected (as analyzed on 12.11, AFTER the election) was 52%, which is clearly wrong.

https://cloud.langfuse.com/project/clv2hvvyw0006z9uchz6u1irw/sessions/DeployableThinkThoroughlyAgent%20-%202024-11-13%2009%3A10%3A45
https://cloud.langfuse.com/project/clv2hvvyw0006z9uchz6u1irw/sessions/DeployableThinkThoroughlyProphetResearchAgent%20-%202024-11-13%2012%3A15%3A58

Current problems to solve:
-> The TTA placed a bet on YES (outcome 0), however agent reported a trade on NO
-> TTA thinks probability of Trump winning, after the election, is only 52%
-> Also consider having the Social Agent follow a different, more reliable agent