Something weird happened yesterday that I can't stop thinking about, even though half the internet is currently losing its mind over new iPhones and I should probably be writing about that instead. (I'm not going to. Everyone else already is, and I ordered mine online at 5:45am my time anyway, so there's not much to report besides "the checkout page was slow.")
No, the thing that's stuck in my head is this quantum computing paper that showed up on a NASA website Thursday night, sat there for a few hours, and then vanished. The Financial Times got a screenshot of it before it disappeared and ran a story saying Google's researchers are claiming they hit "quantum supremacy" - basically that their quantum processor, something called Sycamore, did a specific calculation in about 200 seconds that would supposedly take the world's fastest existing supercomputer around 10,000 years to finish.
Ten thousand years. I keep typing that number out because it doesn't feel real when I just think it. That's longer than written history, for one calculation, versus 200 seconds. Even if the number ends up being generous (and there's already pushback from IBM, who I gather has some stake in not letting that stat stand unchallenged), the gap between "took a couple minutes" and "would outlive our species several times over" is the kind of thing that makes me go quiet for a second.
Nobody at Google is officially confirming any of this yet. The paper wasn't supposed to be public, the working theory floating around this morning is that someone at NASA, who Google partners with on some of this research, posted a draft to their site early by mistake, and it got yanked once someone noticed. So what we've got right now is a leak, a screenshot, and a bunch of physicists on Twitter arguing about what "supremacy" even means in this context and whether it's a useful term at all versus just good marketing. I've seen at least three people point out it doesn't mean the computer is "smarter" than anything, just that for one very specific, kind of contrived problem, it's faster. That distinction matters more than the headlines suggest.
I don't understand quantum computing beyond a surface level, and I want to be upfront about that rather than pretend otherwise for a paragraph and a half. What I do understand is that this is the kind of story that'll get properly explained in a few weeks once there's an actual peer-reviewed paper instead of a screenshot of a screenshot, and by then everyone will have moved on to whatever's next. That's sort of the nature of leaks, they're the most interesting version of a story precisely because nobody's confirmed anything and everyone gets to speculate.
What I keep coming back to is how ordinary the leak itself was. Not a hack, not a whistleblower, not some dramatic breach. Just a file sitting on a public-facing NASA server that somebody forgot wasn't supposed to be there yet, for what sounds like a few hours before someone caught it. The actual content is enormous, arguably one of the bigger physics and computing claims of the decade if it holds up, and the mechanism by which we all found out about it is a filing mistake. I find that funnier than I probably should.
Anyway. If you want my honest take: I'll believe the full "supremacy" claim once there's a real published paper and some independent people have poked at it, not before. IBM's skepticism is worth taking seriously, they're not exactly a disinterested party but they also know this space better than almost anyone alive. Even discounted, even if the real number is way less dramatic than 10,000 years, something is clearly happening in that lab worth paying attention to, and I'd rather write about that today than about which iPhone camera bump looks worse in person.