Every unique game on this drive is an observation. Moves become neurons; recurring contexts become assemblies; openings and motifs become higher bodies. Candidate intentions compete, support one another through learned compatibility, and relax into a coherent next move. Unfamiliar positions drain into the noise label.
The learned hierarchy Neural field Hyper Star ✡ Constructor tasks Neato Burrito CA Lab
Chess transitions DNA complements Three-color difference XOR / parity Toroidal locality Conway's Life · B3/S23 HighLife · B36/S23 Seeds · B2/S Day & Night · B3678/S34678 Brian's Brain Cyclic CA Wireworld Rule 110
Restart view Pulse memory tempo
Layers: whole corpus → opening families → tactical/positional motifs → move neurons. Brightness is current label strength; lines are learned compatibility.
Ask the brain Enter SAN moves separated by spaces. The last three moves select the deepest learned context; shorter contexts provide hierarchical backoff.
Relax intentions
What makes this HRL Objects are contexts. Labels are candidate continuations. Empirical frequency supplies a persistent prior. Shared motifs and suffix contexts supply compatibility. A multiplicative, normalized update repeats until the field settles. The noise label is first-class, and hierarchical backoff lets a coarse memory speak when a fine one has never seen the position.
Constructor-theoretic lens 𝒯 = {input → output}