Does running the generator on the whole fly brain, through every check, find more alloys?

EXPERIMENTS.md · lines 8468–8499

E139 — The whole brain on the whole ladder

164,506 neurons, 25,135,527 edges, against the full ladder — screen, ordering confirm, and the corrected feasibility — restricted to the high-entropy class the project is named for. This is the configuration the project was built to test and it has never been run.

What changed to make it possible: the substrate went from 5,311 mushroom-body neurons to the whole connectome (E137), the reward now carries every correction from E96, E99 and E132 (E133), and out-of-class proposals are refused rather than scored (E134).

The wiring of ladder to brain is unchanged and anatomical: the composition enters at the projection neurons, one element per channel; the Kenyon cells are the sparse expansion; the verdict enters as reward at PAM (appetitive) and PPL (aversive); the score is read from the MBONs, signed by their own measured compartment input. What is new is that signal now propagates through all 164,506 neurons between those points rather than through a type-filtered 2.5 per cent.

Predicted:

1. The whole brain is slower per proposal but not prohibitively — 0.18 s per forward pass measured, against roughly 0.01 s for the 5,311-neuron circuit, so about 18x. A run that took minutes takes hours.
2. Distinct in-class finds per unit spend do not improve much over E136's 5.0. The proposals come from 24 walkers doing run-and-tumble on the learned field, and the walker count is unchanged. Representational capacity was never obviously the binding constraint.
3. The best find is within about 10 meV/atom of CrossEntropy's -78, as at 5,311 neurons the fly reached -74. A larger substrate should not lower the ceiling.

Falsified if distinct finds per unit spend rise by more than a factor of two. That would mean representational capacity was the limit, that the mushroom-body filter had been starving the search all along, and that the whole-brain substrate is the fix. That is the outcome worth having.