And then, simultaneously writing a paper in which he writes that he knows the proof is overly complicated and needs vast simplification before it can be useful and that he knows the timing can invite cynical views but then publishing it anyway in its pre-useful state and under that exact timing which can invite cynical views.
This result turn out to be genuinely amazing in 2-3 years when it is verified and people understand its implications, and then I would be very happy because this is a really cool thing. But until then, it is functionally useless. Nobody can understand or profit from this yet. Given the timing, it's purely for hype and marketing and I'm tired of AI companies doing this.
These days, that line of thinking just gets you scooped by someone who has convinced himself that one fewer nine is good enough. If you have a strong lead on something like this, it's clearly better to put it out there for criticism and analysis earlier rather than later.