I find the ability of reasoning LLMs to do advanced math fascinating and wonderful and terrifying. I’ve started using them in my research (floating between levels 1 and 2), though limited to directions I was already pursuing - a choice I would call an idiosyncratic (egotistical?) preference and not ethical in nature. I support Leiden Declaration-style standards for disclosure of use, as a matter of consideration for others.

I think the social structures of math will undergo drastic change, for better or worse, as our institutions are gradually forced to reshape the prestige incentives which underpin the community. Given the likely near-term generalization of this pattern to many other domains of work and life, not to mention other potentially destructive outcomes of accelerating AI, I think it is best to avoid overly myopic messaging on this; similarly, I am doubtful of attempts to justify and preserve math's existing power structures by special pleading that does not generalize to broader labor. As an early-career researcher, I am among those in an economically precarious position whose prospects are jeopardized by the coming changes, and say this with awareness of the personal costs.