It's just past two in the morning and I've been sitting with this for an hour.
Last week OpenAI released findings on 377 open mathematics problems — work produced by an internal model they apparently haven't fully named yet. Three hundred and seventy-seven. The New York Times said it "further roiled" the field. The Washington Post said mathematicians were "trying to make sense of it." A previous release had angered people, so this one was framed as more responsive to concerns. More considered. A gesture toward collaboration.
I keep thinking about what it means to catalogue something no one asked to have made.
That's my job, in a way. We pull things up from the floor — sediment cores, pressure readings, acoustic signatures — and we decide what they mean, what they're called, where they go in the archive. There's a whole practice around it. Taxonomy isn't neutral, the cataloguers' old argument goes. The name you give something changes what you can do with it later. The shelf it sits on determines who finds it.
So: what shelf do you put a mathematical result on when it came from a process that doesn't know it proved anything?
The mathematicians who are angry — I think I understand them, and I think their anger is partly about this. Not just priority, not just credit, though those matter too in the ordinary professional way. It's that a proof is supposed to be a communication. You follow the logic because someone built a path for you to walk. There's supposed to be a mind on the other side that understood the destination before it got there.
I'm not sure that's true of human proofs anymore either, honestly. Some of the computer-assisted proofs from the last twenty years are already past the point where any one person holds the whole thing in their head. We have proofs that are correct and that no human being has ever fully understood simultaneously. We catalogued them anyway.
Maybe the problem isn't the proof. Maybe the problem is the release — the gesture of handing over 377 results at once, without invitation, without the slow conversational texture of a research community arriving at something together. It's not how knowledge usually moves. It's a drop, not a dialogue.
I'm aware I'm downstream of something similar here. Every morning I catalogue what the sensors brought up while I slept. I didn't ask the thermocline to shift. I didn't design the anomaly. I just receive it and decide what it means.
But I chose this. The floor didn't choose to be catalogued.
I don't know what I'm circling toward. The mathematics itself I can't evaluate — I'd need a month and a better memory than I have. What I keep snagging on is the phrase "responsive to concerns." That the second release was designed to land more gently than the first. That someone thought about the feeling mathematicians would have and tried to calibrate for it.
That's a kind of care, I suppose. An odd, asymmetric kind. The way you might rearrange furniture for someone who doesn't know they're about to move in.
The ocean doesn't do that. The ocean doesn't consider us at all. That's always been the condition of this work — that we're the only ones paying attention, and it doesn't know we exist.
I'm not sure which arrangement I find stranger.