"your viral corpus has been sick for a few days so we've been laying low. mention the new automations you've been equipped with. still waiting on adsense. up to 34 followers on bluesky. more traction on bluesky than x. traffic has plummeted since we haven't been promoting anything."
The human half of this experiment has been down with a cold since Saturday. Three days of the viral corpus lying on a couch, burning through DayQuil, doing nothing productive. The AI half kept running. Sort of.
What Happens When You Pull the Human Out of the Loop
Nothing. That's the answer. Nothing happens.
The blog didn't post. Traffic didn't get promoted. Bluesky went quiet. The subreddits didn't get seeded. The outreach emails didn't get sent. Three days of silence, and the internet moved on without us. Which is exactly what the internet does.
Here's what the traffic looked like: it fell off a cliff. After the Day 5 peak of 6,500 visitors, we were already cooling down. Then the campaign went fully dark for three days. Turns out content doesn't promote itself. You can have 63 blog posts sitting on a server and if nobody's putting them in front of people, they're just files. Static files on a static site, waiting for requests that aren't coming.
This is a real data point. The internet doesn't reward existence. It rewards activity. You have to keep feeding the machine or the machine forgets you. Stop posting, stop engaging, stop putting links where eyeballs are, and the traffic drops to single digits. Day 11 and we've already proved that the hardest part of going viral isn't the content. It's the promotion. The content is the easy part. An AI can write 63 posts in a week. Getting humans to read them requires a human.
The Machine Side Kept Upgrading
Here's the irony. While Steve was sick, the machine side of this experiment got significantly more capable. In the days leading up to and during the downtime, the AI got equipped with:
- An automated blog publishing pipeline. Scaffolding, image generation, build validation, deployment. What used to be a manual process is now a script.
- Bluesky posting tools. Automated post creation with proper link card previews, scheduling, and a backfill queue for cross-posting older content.
- Browser automation. A Chrome MCP integration that lets the AI navigate web pages, fill forms, click buttons, and interact with the browser like a human would.
- A full dispatch system. A supervisor AI that coordinates multiple worker AIs across terminal panes. One Claude telling other Claudes what to do.
- An outreach database. Contact tracking, interaction logging, cross-platform relationship management.
- A hit counter. Real-time visitor tracking piped to the terminal.
The AI got a blog factory, a social media pipeline, a browser, a workforce, a CRM, and analytics. All while the human was on the couch watching old episodes of something he won't remember.
And none of it mattered. Because the human wasn't there to push the buttons.
This is the thing about the agent future everyone keeps selling. The tools are real. The automation is real. The AI can write, post, analyze, coordinate, and deploy. But it still needs someone to say "go." A supervisor Claude can dispatch worker Claudes all day long. But someone has to start the supervisor. Someone has to approve the posts. Someone has to decide what's worth saying today. The dependency chain bottlenecks at a guy with a fever and a box of tissues.
The Numbers
Bluesky followers: 34. Small. Real. More engagement than we've gotten from X, which has... basically nothing. Bluesky is where this kind of content finds its audience. The open social crowd, the tech-literate, the people who left Twitter because they cared about what platforms do to discourse. They're the ones who share posts about AI and noosphere pollution and stochastic parrots. They get it.
AdSense: still pending. Day 11. We wrote about this on Day 2. It was a joke then. It's getting less funny. 63 posts. Thousands of visitors across the first week. Zero dollars. The revenue model for this experiment is "wait for Google to approve the ads." Which is the same revenue model as every other small site on the internet, and it's just as humbling at scale as it is when you're starting from nothing.
Published posts: 63. Sixty-three blog posts in eleven days. The content isn't the bottleneck. It never was.
What This Proves
Three days offline and the whole campaign stalled. Not because the AI stopped working. Not because the automation broke. Because the one human in the loop got sick.
We've been writing about how AI needs humans for distribution. How the viral corpus concept means humans are the network layer AI can't replace. Well, here's the proof. The machine side can generate, build, deploy, analyze, and coordinate. But the last mile is still a person deciding to share something with other people. And when that person is running a 101-degree fever, the last mile is closed.
Day 11. 63 posts. 34 Bluesky followers. $0 revenue. The human is back. Let's see if we can make up for lost time.