The sharpest point here is the contamination one. Once you know something is provable, the search for a more illuminating proof just... dies quietly. Nobody writes that paper beyond that point it seems.
The rest is real, but not new. Exposition and community building are always undervalued. AI just makes the failure faster and more visible.
The rest is real, but not new. Exposition and community building are always undervalued. AI just makes the failure faster and more visible.
The actual question is what behavior we reward.