Cross-domain transfer learning in a minimal Hebbian cortical network: from Tic-Tac-Toe to Connect Four without retraining

We present Brain over Brains aka BoB, a purely Hebbian learning system (no backpropagation, no reward shaping) organized as any cortical columns needed of any neurons required for each (the current publication covers 3 columns of 5 neurons so 15 neurons total). After observing 18 games of Tic‑Tac‑Toe (3×3 grid) : After each run of 4 Tic-Tac-Toe grid, BoB is exposed to a never seen brand new grid, BoB was able to solve the position, Then the system was exposed to Connect Four (6×7 grid) without any retraining or parameter adaptation. The same 15 neurons activated on Connect Four (100% overlap), 134 new causal path signatures were sedimented in 66 seconds, and chains of up to 4 neurons in cascade emerged. The system successfully played a winning move on a never‑seen Connect Four board requiring immediate‑win recognition.