Blog
On the day the emergency was reported, my entire contribution was telling our AI what I had seen in the news and answering “yes”.
On 21 October 2026, two vulnerabilities were found in PHP that let an attacker run code on the server without authenticating, and emergency fixes were released (CVE-2026-109618 and CVE-2026-109619, severity 9.2 out of 10). Every environment where PHP handles web requests was affected, and because it can be triggered before any application code runs, it mattered regardless of what you had built.
Seeing the news, the first thing I did was ask our AI.
That was all I asked. I did not look up the CVE numbers and hand them over, and I did not specify which machines to check. It is the sentence anyone says after seeing the news.
What came back was neither “you are fine” nor “you are in danger.” It was what in our setup could be affected and in what order to deal with it — a list, prioritised.
After that it asks, item by item, “shall I do this?” and I answer yes. Each answer moves the work along.
As it happens we do not use much PHP. Even so — investigating, prioritising, working through the list, confirming it was done — all of it finished inside tens of minutes.
Compared with the days when I administered servers myself, the difference is night and day. Time used to evaporate in the gathering stage. Which machine runs what, again? And once you remember, you are back to looking up the procedure. The whole time there is a low hum of “what if they are already in.” A weekend disappears.
What actually wears you down in an emergency is not the work itself. It is the investigating, the judging and the sequencing. Hand that part over and what remains is saying yes.
To be clear: the AI did not go off and fix things on its own. Every action is put to me first, and only what I approve gets done. If a change looks significant, that is the moment to stop and think about it.
My own contribution was repeating what I had seen in the news, and answering “yes”. If that is all it takes, then on a day like that there are better things for a company's owner to be doing.
We run our own company on an AI server we built ourselves, miraiGen.™Cube. We grew it in-house, found it good, and that is why we turned it into a product. Even now that we sell it, days like this keep showing us why.
On the day an emergency vulnerability landed, we did not have to panic. That alone made me glad we built miraiGen.™Cube.