Episode 2 · The tech
Episode 2: The Growth Engine That Runs Itself
Most review sites hide how they are made. I am going to show you the machine room, because the automation is half the reason I think this 12-week bet is even possible with a team of basically one.
Here is the line that matters, up front. The agents run the operations. Humans run the judgment. Nothing about which tool wins, or what breaks in week three, is decided by a script.
What the machine actually does
Every Monday, without me touching anything, a set of agents wakes up and does the boring, repeatable work that usually eats a solo operator’s whole week.
It pulls the real Search Console numbers, this week versus last, and flags which pages moved. It finds pages that just started showing up in Google and tells me where they rank. It writes the week’s entry on our public changelog, checks every link resolves, and publishes it.
It even drafts this story’s data. The US click count you see on each episode is pulled straight from Search Console by a script, not typed in by me hoping you will not check.
Why automate the operations and not the reviews
The scarce thing on a review site is not words. It is judgment. Whether a CRM falls apart at 50 users, whether a tool’s “free” tier is a trap, whether the pricing page is lying to you. That takes an operator who has felt the pain.
Words are cheap. Judgment is not. So the automation exists to clear the operator’s desk of everything that is not judgment. Reporting, indexing, distribution, the changelog, the freshness checks. All of it runs on rails.
That is the whole design. Let the machine do the treadmill so the human has time to make the calls that actually decide whether a review is worth reading.
The guardrails
Automation on a content site has one failure mode, and it is slop. Pages that read like they were assembled, not written. Google spent all of 2025 and 2026 learning to smell exactly that, and it demoted us for a version of it once already.
So the rules are strict. A named human signs every ranking and every verdict. The ops agents touch data and plumbing, never opinions. And when the changelog auto-publishes, it runs through checks first: no broken links, no duplicate entries, a real build that has to pass before anything ships.
The machine is fast. The human is the brake. That combination is the whole strategy for the next 11 weeks, and you are going to watch it play out here, with the real numbers, every Monday.