Between the week of March 9 and March 17, 2026, three WordPress publications became one platform. This is what happened in those two weeks, with the numbers.
The archive moved first. Every article and page from The Farmington Mercury, The Charlotte Mercury and Strolling Ballantyne went into one database: 830 posts and 126 pages, 956 records, validated on March 15 with zero orphans. Farmington alone was 524 articles. The Charlotte Mercury brought 82 pages of guides and hubs with it. Nothing was left behind and nothing was retyped.
Then the images. 7,124 of them, pulled off three WordPress libraries and put into one storage system with permanent addresses: 2,334 for Charlotte, 4,288 for Farmington, 502 for Ballantyne. Every article's original photograph came with it. No placeholders.
Then one site to serve all three. A single codebase that reads the domain a reader typed and serves the right publication. Type cltmercury.com and you get The Charlotte Mercury. Type strollingballantyne.com and the same system gives you Strolling Ballantyne. One design, three expressions: a lead story, a grid of secondary stories, beat navigation organized the way a city desk is organized, author pages, and articles that load in under a second from a global network instead of one shared hosting account.
Then the cleanup nobody sees. About 550 broken links repaired. About 60 redirect rules so that every old address still lands somewhere. Every publication registered with search engines under its own name. Read-and-write permissions tightened so the public can read everything and change nothing.
Then Mercury Local itself. On March 17 this site and my personal site came off the rented app builder and onto the platform, 23 posts and 3 pages, with the domains cut over the same day.
By March 16 the count was 813 published articles across three publications, six production domains, one codebase, one database, one deployment. Zero broken image addresses. Zero WordPress dependencies. Three WordPress installations and three hosting accounts retired.
Who did it
One publisher and an AI. I want to be precise about that, because it is the part people ask about and the part that is easiest to overstate.
I decided what the system was: one database, publications as rows in it, the domain as the only thing that tells the site which publication to be. I decided what got kept (everything) and what got fixed on the way through (the links, the redirects, the permissions). The AI wrote the migration, the site, the image transfer and the redirect rules, and ran them. When something came back wrong, I said what was wrong and it fixed it.
That division of labour is the subject of the last piece in this series, so I will leave it there for now. The short version: the architecture was mine. The engineering was the machine's. Two weeks.
Why two weeks matters
Not because fast is good. Because it changed what a publication costs to start. On WordPress, a fourth publication was a fourth site, a fourth hosting account, a fourth theme to fight. On this platform it is a row in a table. Two weeks after these two weeks, we launched Strolling Firethorne from nothing in a day. By September there were sixteen.
That is not a productivity story. It is a different cost structure, and a different kind of company follows from it. The next fourteen pieces are about the parts.
This is No. 02 in We Built Our Own. Previous: Three WordPress blogs and a rented app. Next: The day we deleted the media server.