Source: notes from closed PRs #238/#239/#281.
Across the 06-10/06-11 cycles the actual fitness wins came from transfer_score, while the stated rationale (driving consciousness/phi toward 0.281) was a weak predictor of the real gains (e.g. #281: transfer +0.028 vs consciousness +0.0007).
Two related items:
- Optimize directly against transfer_score (A-phase attractor geometry, alpha sweep) instead of via the consciousness-target proxy.
query_gravity is stuck at 0.373 and currently instrumentation-only. Decide: fold it into the fitness function, or boost it via stronger drive / higher order parameter R — right now it's a dead metric.
Source: notes from closed PRs #238/#239/#281.
Across the 06-10/06-11 cycles the actual fitness wins came from transfer_score, while the stated rationale (driving consciousness/phi toward 0.281) was a weak predictor of the real gains (e.g. #281: transfer +0.028 vs consciousness +0.0007).
Two related items:
query_gravityis stuck at 0.373 and currently instrumentation-only. Decide: fold it into the fitness function, or boost it via stronger drive / higher order parameter R — right now it's a dead metric.