Tested K&A at scale, found something, need your help

[oshfdk]

What would be your best advice or best direction for a computational analysis by Machine Learning (not hallucinating LLM) ?

I think it makes sense first to spend a few weeks or more studying what has been attempted so far. For some recent ML attempts I think there have been a few threads by @quimqu.

Personally, I don't think ML is of much use without some good underlying hypothesis that can explain many puzzling features of the manuscript. I believe most attempts of just throwing some computational power at the manuscript and expecting a breakthrough are doomed.