Everyone has benefited from this effort even if we don't have one spec at the end.
The intense effort to find common ground, spot irreducible conflicts, to understand the other spec, and port over good concepts, resulted in both EIP-8130 and EIP-8141 being much better.
Actual neural net being trained in the EVM. 200K gas inference, about 340K gas training step, including loading and storing all weights to storage with MIP-8. Still room to optimize.