Placeholder writeup. Layer 1 (steering) is built and playable; layers 2 to 5 are planned. The written breakdown is a draft and is not yet published.
The five breakthroughs from Bennett's history of intelligence, rebuilt as stacked reinforcement-learning agents in one shared foraging world. One layer per breakthrough, each one on top of the last.
Placeholder writeup. Layer 1 (steering) is built and playable; layers 2 to 5 are planned. The written breakdown is a draft and is not yet published.