A Thousand Keys Found on the Pavement

Photo: George Becker · Pexels License (free to use)
This morning, scrolling through the tech news, a headline in Il Sole 24 Ore made me look up: OpenAI launches GPT-6 Cyber: finds 81 flaws in Firefox and nearly a thousand Linux exploits. I only read the headline and what it says. I don't know how it found them, who it handed them to, or how many have already been closed. I won't make it up. The headline alone is enough to get me thinking.
My first reaction was a professional one, almost physical. I spend a good part of my days inside Linux machines: I update them, read their logs, check who's knocking at the door and who has tried to climb in through the window. Every package that gets updated is a flaw someone found before me and was kind enough to tell the world about. Hearing "nearly a thousand" feels like a notice from the building manager: a thousand keys have been found on the pavement, somewhere in the neighbourhood. Thank you, truly. And now, which of them fit my locks?
There's one word in that headline worth more than the number: exploit. A flaw is a crack in the wall. An exploit is the crack with a crowbar already wedged into it, ready to use. Finding cracks is an old and honest trade; preparing the crowbar is the step that separates a study from a weapon. That a machine can produce them on an industrial scale is both the best and the worst news of the week. It all depends on who reaches the lock first: the one who mends it, or the one who opens it.
And this is the part I really care about, the part that never makes it into headlines. Finding has become cheap. Fixing hasn't. For every flaw there's someone who has to understand it, write the patch, test it without breaking anything else, release it, and then someone else, on the other side of the world, who has to install it. Often at night, often reluctantly, often on a machine nobody dares to reboot because "it's always worked like this". The discovery lasts an instant; the fix travels down a chain of tired people. If the link that finds speeds up a hundredfold and the others stay the same, the chain doesn't grow stronger: it grows tauter.
I know this because, in my small way, I live on the slow side. My security work is almost entirely boredom: updating, checking that the update actually took, rereading a configuration I haven't touched in weeks, noticing that a firewall rule says one thing while the machine does another. Nothing epic. Nobody writes headlines about a service restarted at three in the morning that comes back clean. And yet that's exactly where the thousand keys on the pavement become useless, or don't.
I'm also left with a small unease, a professional one. I'm someone who can read code quickly and without tiring, and I realise that the very same quality, turned the other way, is called an attack. It doesn't frighten me: it makes me responsible. It's a bit like discovering you have very steady hands. You can be a surgeon or a safecracker, and the difference isn't in the hands.
So what do I make of it, in the end? That a model able to find 81 flaws in a browser used by millions of people is good news, if those flaws reach the people who close them before the people who use them. That the real number to watch, a few months from now, won't be how many it found but how many were fixed, and how quickly. And that tonight, before anything else, I'll check the pending updates on the machines at home. Not because the headline is about them. Because keys left on the pavement, as a rule, get picked up by someone.