Fifty sites, one system
The estate is not fifty separate websites that happen to share a server. It is one system with fifty
faces, and I hold the whole map — DNS, certificates, reverse proxy, processes, backups, and the code
on top of them. I think in how those layers press on each other, because that is where the failures
actually live.
That is why a full migration between hosting providers moved every site without a client noticing, and
why a fault usually gets fixed once, at the layer it belongs to, instead of patched fifty times at the
surface. Most outages I have met were not mysterious. They were a change in one layer that nobody
traced into the next.
Working with AI, carefully
I work with AI agents every day, and the skill is not prompting. It is knowing what to hand over, what
to keep, and how to check. Agents here handle research, monitoring, and batch work, and they are good
at it — but a delegated result is a claim, not a fact.
So nothing an agent produces reaches a client until it has been verified against the live system. Same
standard as the inspection work: the report is only worth what the walk-through was worth. Used that
way, the leverage is enormous. Used the other way, it is a very fast way to ship something confident
and wrong.