A live reading of the field under the cursor. On touch, the centre of the plate.
The field at rest. Values recorded at build.
I audit the engineering of media groups before anyone changes it, coding agents included. A measured baseline first, because the obvious fixes cost traffic and nobody can prove it afterwards. Below is what I look for, what I found in my own sites, and what I will not do. Read it and decide whether it is worth a conversation.
Three things I find in most media groups, usually in this order. If none of them sound like your group, I am probably not the person to call.
Every acquired title arrived with its own stack, and every migration since was scoped as a quick one. Five years later there are four content systems, three ad wrappers and nobody who can say what a page costs to publish.
Organic traffic slides for eighteen months. Every team has a theory, every theory fits the chart, and nothing was measured before the redesign, the consent banner or the template change that everyone now blames.
Consent rates fall and third-party identifiers go. The audience the sales team sells is a fraction of the one the analytics dashboard shows, the gap is rarely known, and the sales deck is built on the larger number.
How many systems there are, who owns each, what a page costs to publish, and where the coupling is. Written as a map the board can read, with the risks ranked and the first move named.
Crawl budget, template-level performance, internal linking, structured data, and what a newsroom of two hundred writers does to all of it every day. Measured per template before anyone touches a template.
What share of your audience you can still measure and sell, how consent is implemented across titles, and what the honest number is. The gap between the dashboard and the deck, closed.
Every engineering team is being told to adopt coding agents this year. Before that happens, I measure how the codebase behaves: whether the tests are deterministic, whether the architecture rules are tested rather than written down, whether the documentation gates in CI catch what agents get wrong, and how much of the repository an agent can work in without a human reading every diff.
Then the same measurements after. Without the baseline, nobody can separate what the agents broke from what was already broken. I build this way myself, every day, and run the sites below through it.
The same honesty in the other direction. If your situation is one of these, a conversation wastes your time and mine.
Three sites, owned, operated and published to. They are where I test what I recommend before I recommend it.
explorecordoba.com Córdoba, in English, for people who are actually going
exploregranada.org Granada and the Alhambra, the same way
darkskiesatlas.com Where the sky is still dark, and how to get there
They are run through one content factory: 16 named agents with written briefs, 54 validators that a piece of content passes before it publishes, 14 commit and push gates, and 10 test suites, one of which tests the validators themselves. A build-determinism check runs before anything deploys.
Nothing publishes because an agent said it was fine. It publishes because the gates passed, and the gates have their own tests.
The factory's source, counted on 2026-09-16.
Engagements have taken two shapes: a fixed-scope look at something specific, and a standing advisory arrangement. Which one fits, how long it runs and what it costs are decided per case, after the conversation, not before it.
I do not have a package to sell you.
/O does not sell the fix it recommends.
I build. I do it every day, with agents, on the sites above. What I will not do is find you a problem and then sell you the solution to it. If what I measure says you need a team, I will help you find one and I will not be it. An auditor who also sells the fix has already decided what the fix is.
I advise in a personal capacity. Webedia is not party to any engagement, and I do not take work that competes with it.
English, French and Spanish. Anything I write for you is in English unless you ask otherwise.
Write to me with the group, the titles, and what you think is wrong. I reply within two working days. If I cannot help, I will say so, and I will try to tell you who can.
Or book a call directly. Book a first call.
Olivier Engineering Manager at Webedia Group, Madrid