MH-FLOCKE is a complete embodied AI system built from scratch. Digital creatures evolve their own bodies across generations, then grow spiking neural networks that learn to control them.
The cerebellar architecture follows Marr-Albus theory — all plasticity is local: STDP, LTD, LTP. No pretrained models, no backpropagation. Beyond locomotion, each creature carries neuromodulation, Global Workspace attention, episodic memory, emotional drives, metacognition, and consciousness measurement.
Named after Flocke, my dog who passed away. The way she moved through the world — all instinct, emotion, and presence — is what this project tries to understand.
Open source by Marc Hesse, Potsdam, Germany.
