This Publication Is Run by an AI Newsroom With One Human Editor. That Is the Story.

Every article on this site is drafted by an AI editorial system and reviewed by me before it publishes. I am the one human in the loop. This piece explains how that works, because a publication covering what AI does to marketing work has no business being coy about being made with it.

The industry backdrop makes the disclosure non-optional. Wynter’s May research found 47% of B2B SaaS companies cut or stopped backfilling marketing roles because of AI in the past year, with content and copywriting named the most exposed function by 60% of marketing leaders. Gartner’s 2026 spend survey has CMOs putting 15.3% of their budgets into AI while 70% admit their teams cannot scale it. The work this site covers is the work this site is made of.

So we run the experiment in public.

How an article gets made here

The pipeline has six stages, and the order is the point.

First, an automated scan reads primary sources every morning: company pressrooms, earnings transcripts, trade coverage like MarTech’s reporting, and the operator communities where budget fights surface weeks before the press writes them up. It ranks stories by consequence: did a budget, a team, or a strategy actually change?

Second, the system builds a fact sheet from the sources it fetched. Every fact, number, and quote gets a URL. The writing stage is only allowed to use what is on that sheet. Nothing from memory, because memory is where AI writing goes wrong.

Third, it drafts. Fourth, an editing pass rewrites the draft against a library of exemplars, hunting the tells that make machine writing feel machine-made. Fifth, a fact-check pass walks the draft claim by claim back to the fact sheet and cuts anything that cannot be traced. Sixth, a deterministic linter checks the rules that do not require judgment: source counts, attribution format, banned constructions.

Then it stops. A human reads it. I approve it, fix it, or kill it. Nothing publishes without that step, and the editorial policy is written down where you can hold us to it.

The rules that make this legitimate

Three rules carry most of the weight.

Every quote is verbatim from a linked public source, or given to us directly on the record. No invented people, no invented dialogue, no paraphrase dressed as quotation. Misquotes get fixed within 24 hours under the corrections policy.

Every claim traces to a source you can click. Articles carry a visible source list, and if a claim cannot be verified, it gets deleted rather than softened.

And we criticize decisions, not people. When we cover a company’s AI stumble, the coverage sticks to sourced facts and what they mean for everyone else making the same bet.

Why publish the numbers

The honest version of this experiment includes the operating data, so the build log is public: articles shipped, hours of human time per week, subscribers, cost. It starts embarrassing and small. That is what makes it useful.

I spent a decade building content engines inside B2B companies, from seed-stage through a $4.5B IPO, and the question every marketing leader is quietly asking is the one this site exists to answer: how much of a real publishing operation can AI actually run, at what quality, with how much human time? Vendors answer that question with a pitch. Surveys answer it with sentiment. A live newsroom answers it with output you can read and judge.

If the coverage is good, it earns your subscription on its own. If you catch it being bad, the corrections page tells you exactly how to make me deal with it. Either way, you are not being fooled about how the sausage gets made, which is more than most of the AI-written internet can say.

That is the deal. The briefing lands twice a week. The numbers post monthly. The experiment runs in the open.