I once bought the domain insulatedplasterboard.com for $10.
This was around 2010. I'd figured out that exact-match domains for keywords with decent search volume got practically handed a first-page Google ranking. Insulated plasterboard had about 2,000 searches a month. Not sexy. Didn't need to be.
I wrote one article (and I use the word "wrote" generously). Spun content, stitched together from other sources, reorganised just enough to look original. AdSense ads designed to blend into the page so they looked like navigation. The goal wasn't to inform anyone about insulated plasterboard. The goal was to get a human to land on the page, mistake an ad for content, and click.
Ranked first within two weeks. Two to three dollars a day. I had dozens of sites like this. The portfolio added up.
Then one morning Google banned my AdSense account. Years of income, gone between breakfast and lunch. No appeal. Which was fair, I'd been violating the spirit of the entire system since day one.
I didn't think of myself as someone producing content for machines. I thought I was gaming a system. It took years to recognise what I was actually doing: writing for an algorithm, with humans as the revenue byproduct. The person who needed information about insulated plasterboard was never my audience. Google was. The human was just the thing that carried a wallet past my billboard.
That was over 15 years ago. I was producing content for machines, not people. I just didn't have the vocabulary for it yet.
The Metric Becomes the Game
There's a concept called Goodhart's Law: when a measure becomes a target, it ceases to be a good measure. I didn't learn it from a textbook. I lived it across three career phases without ever naming it.
Phase one: SEO. The measure was Google rankings. So I optimised for rankings. Content quality was irrelevant, what mattered was keyword density, backlink profiles, domain authority signals. The human reader was already a byproduct.
Phase two: content businesses. After Google killed my AdSense empire, I moved into operations for online businesses. email infrastructure, subscriber journeys, conversion funnels for a language learning blog with a million monthly visitors. The engine worked beautifully. Open rates climbed. Click rates climbed. Conversion rates climbed.
But the metric was never "did this person learn a language." The metric was "did this person buy the thing." Those are not the same question. The system didn't require anyone to ask whether what we were converting people into was worth their money. It required conversion rates. So that's what we optimised for. Not because we were evil. Because that's what Goodhart's Law does: it makes you so focused on the proxy that you forget what it was supposed to measure.
Phase three: now. I'm a systems architect. I clean up the aftermath of the first two phases: mthe duct tape empires, the tag chaos, the zombie automations, the subscriber journeys nobody designed intentionally.
But even in this work, I see the same pattern. The metric is engagement. I'm building cleaner machines now. Better machines. But they're still machines, and the human is still being moved through a system.
Frederick Taylor put a stopwatch on factory workers in 1911 to find the One Best Way to perform every task. "In the past the man has been first," he wrote. "In the future the system must be first."
When I look at a well-built email automation sequence today, I see Taylor with a stopwatch. A system designed to move a human being through a predictable path like a component on an assembly line. I know this because I'm the one who calibrated the path.
We spent the last century Taylorising physical labour. We've spent the last 15 years Taylorising digital attention. And I've been holding the stopwatch the whole time.
The Loop Closes
Here's the trajectory I didn't see while I was inside it:
- Humans make, humans consume. The old internet. Pre-2010. Someone writes something, someone else reads it. Quaint.
- Humans make more easily, humans consume. The content era. WordPress, YouTube, podcasting tools dropped the barrier to zero. What followed was predictable: a flood of mediocre content, platform algorithms becoming kingmakers, the middle tier getting crushed.
- AI makes, humans consume. 2023 to now. And here's what gets me -- my spun articles on insulatedplasterboard.com were garbage. Barely coherent. And they worked well enough to make money for years. If my garbage was enough to game the system in 2010, what does genuinely readable AI content do to it in 2025?
- AI makes, AI consumes. Emerging now. AI agents reading AI-generated content. Summarising it for other agents. Optimising pages for AI crawlers instead of human readers. The human isn't even the byproduct anymore: they're absent from both ends.
Before the power loom, weaving was a skilled middle-class profession. The machine didn't just make weaving faster, it made average weaving free.
The middle-tier weavers got wiped out overnight. The market split into mass-produced slop and bespoke tailoring for the elite.
That's exactly where content is heading right now. But even that framing is too narrow: it assumes there's still a human at the end who wants something. Step 4 breaks the model entirely. If agents are generating for agents, optimising for agents, summarising for agents, what is the internet even for?
The Ground Shifts
My entire positioning is built on designing intentional human journeys through an online business. The path a real person takes from stranger to customer to advocate. That's my thing.
If the humans leave the loop, what am I architecting?
The fear that sits heaviest isn't "will AI take my job" in the crude sense. It's subtler and worse: do I pivot to building for agents instead of humans, and in doing so become the thing I've spent three career phases slowly walking away from? Do I pick up Taylor's stopwatch again, except this time I'm timing machines pretending to be humans interacting with machines pretending to be helpful?
The Fortress Problem
I see two paths becoming parallel economies.
The agentic economy. Build for AI agents, AI crawlers, AI summarisers. This is where the scale money goes. The people who built SEO empires in 2010 will build agent-optimised empires in 2027. Same playbook, different players.
Human fortresses. Walled gardens explicitly for humans. No AI in the loop on either side. Human creation, human consumption, human interaction. The premium, scarce, authentic layer.
My instinct says fortresses. My gut says that's where the work I want to do lives. Human connection, human trust, human judgment: things I believe in, things worth spending a career on.
But here's the trap, and I can see it clearly because I used to be the trap.
The moment "human-made" becomes a premium signal, someone will fake it. Period.
The moment "authentic" is the metric, optimisation for "authentic" begins. Someone builds the insulatedplasterboard.com of authenticity: designed to look real, engineered to pass whatever filter exists, zero substance underneath. I know this because I was the person who gamed those spaces. I know exactly how that playbook works because I ran it.
The fortress has a Goodhart's Law problem of its own. "Authentic" becomes the measure. The measure becomes the target. And then it ceases to be authentic.
There is no clean escape from this. The same dynamic that ate content, that's eating the entire internet: it follows you into the fortress. Because the fortress is just another system, and systems get optimised, and optimisation eventually hollows out whatever was valuable about the thing being optimised.
The Honest Bet
Right now I'm building the closest thing to the opposite of what I spent 15 years building. Clean systems instead of duct tape. Visibility instead of manipulation. Subscriber experiences designed for actual humans instead of conversion funnels designed for dashboards.
I tell myself this is different. And it is different. The intent is different. The outcomes are better. The humans at the end are having a better experience.
But it's still a system. It's still automation at scale. I'm still the guy with Taylor's stopwatch, I'm just holding it more gently and asking nicer questions about what to measure.
Maybe that's the honest answer. Not fortresses or agents. Not a clean philosophical resolution. Just: do the work, hold the contradiction honestly, and stop pretending it resolves into something neat.
Every business needs to grow or it dies. Growth requires scale. Scale requires systems. Systems require reducing human complexity to manageable inputs. The Taylorism never actually stops. You just get better at not staring directly at it.
The content era wasn't a chapter. It was the prologue. We thought we were democratising creation. We were training the replacement. And I was one of the trainers.
I just build cleaner systems now.
I think.