RLFH is not a panacea!
200-300 < 1000 ~ SFT
RLs Game the the the the 
Newer and better models
Developing more sophisticated agents
Memory { 1) Episodic, Semantic, and Procedural memory }