Until They Come and Get Us
What Mercor, the SpaceX IPO, and a resigned AI safety head tell us about where this is actually headed.
This morning, SpaceX started trading on the Nasdaq.
Elon Musk — the man who said he’d put data centers in space so AI companies never have to worry about compute again — is now officially selling shares to the public. To us. To the same people whose jobs the AI those data centers power is quietly making redundant.
I brought this up with Nidhin in Episode 11 and he laughed. Not because it’s funny. Because it’s so perfectly circular that laughing is the only reasonable response.
“We are the suckers who are buying into that,” I said, “with the money that we earn from the jobs that they want to replace. With our investment into what they built, that in turn screws us.”
That’s not cynicism. That’s just the logic of the system we’re all living inside right now. And nowhere is that logic more visible — or more brazen — than in a company called Mercor.
“Organizing Human Intelligence to Power the AI Economy”
That’s Mercor’s actual tagline. Go look it up: mercor.com.
Read it again slowly.
They are, by their own description, organizing human intelligence — your skills, your expertise, your years of hard-won knowledge — to power an economy being built to need you less and less.
Three 21-year-olds dropped out of their universities in 2023, built this idea, and by 2025 had raised $350 million in a Series C round backed by Peter Thiel, Jack Dorsey, and Larry Summers. They grew from $1 million to $100 million in revenue in eleven months. A Benchmark partner reportedly took a helicopter just to get to the meeting on time.
Everyone could see what this was. Everyone said yes anyway.
What Mercor Actually Does
Here’s the part I want to lay out cleanly, because I don’t think I explained it well enough in the episode.
The AI models built by OpenAI, Meta, and Anthropic have read everything. Every book, every paper, every website. They have the knowledge. What they don’t have yet is the skill. The judgment. The thing that comes from years of actually doing the work.
That’s where Mercor comes in.
You apply, upload your resume, and sit for a one-way video interview — conducted by an AI. If your expertise matches what a client needs, you get offered contract work at $50 to $150 an hour. Your job is to interact with AI agents, correct their outputs, guide their reasoning, and in doing so — transfer your professional expertise directly into the model.
The clients paying for this? Meta. OpenAI. Anthropic.
You are not just doing gig work. You are the curriculum.
As I put it in the episode: “The wisdom we’ve gained from years of experience — from getting out of college to now — what we have learned, the experience that makes us valuable in the industry... that’s what we are now getting hired, paid, to transfer over.”
And the kicker? Nidhin saw exactly where this ends.
“At some point, Mercor becomes redundant. Because AI will have absorbed everything it needed to learn. And then it will say — I don’t need you anymore.”
Then It All Blew Up
In April 2026, Mercor confirmed it was among thousands of companies hit by a supply chain attack through a compromised open-source tool called LiteLLM — widely used across the AI industry. The extortion group Lapsus$ claimed responsibility and published samples of stolen data including internal Slack communications, ticketing logs, and recorded interactions between Mercor’s AI systems and its contractors.
The breach exposed what was really being collected: Social Security numbers. Passport data. Facial biometrics. Voice recordings. Video interviews. Screenshots of contractors’ computers, taken silently at random intervals by monitoring software they were required to install.
At least seven class-action lawsuits followed.
One of them alleged something particularly striking: contractors were being encouraged to use real company data in their training tasks — proprietary information from their actual day jobs — provided it was slightly altered. When one contractor tried to avoid including sensitive details, reviewers pushed back and said the work was too vague.
So not only were people transferring their own expertise into AI models. They were potentially transferring their employers’ expertise too. Without anyone’s knowledge or consent.
Meta paused all contracts with Mercor. OpenAI began its own review.
The Part Nobody’s Talking About
Here’s what stayed with me after the episode.
Mercor’s biggest clients — the companies most alarmed by the breach — weren’t primarily worried about the contractors. They were worried about losing their competitive edge. About their AI training methodologies being exposed to rival labs.
The workers were secondary. As always.
One contractor, after Mercor abruptly shut down a project and then offered to rehire the same people at a lower rate, said: “We are real people who deserve some notice, or warning, or some consideration. I know we are working with AI but we don’t work for AI.”
But here’s the uncomfortable truth — in a sense, they do. We all kind of do.
“The ‘We’ Is Important Here”
Nidhin made a distinction in the episode that I keep coming back to.
When we ask “are we enabling AI to replace us?” — the we matters.
The people teaching their skills to Mercor aren’t the ones pulling the strings. They’re responding to an economy that has narrowed the options. Jobs are scarce. The gig economy is what’s available. You take the work you can get.
“People like you and me — I don’t think we count in that we,” Nidhin said. “If we were smart enough, we’d be on the other side of Mercor.”
The we that is actually enabling this — deliberately, strategically, with full knowledge of what’s being built — is a much smaller group. The investors. The lab executives. The founders who took the helicopter ride.
And then there’s the third we. The one that hit hardest.
Us. Buying the SpaceX IPO today. Funding the next round. Signing off, in our own small way, on the whole thing with money earned from the jobs that are quietly disappearing.
It’s not a conspiracy. It’s just the logic of the system.
What Comes After the Jobs
This is the question Nidhin and I kept circling but never fully landed on: what actually happens after the displacement?
The optimistic version — the one industry leaders keep offering — is that new jobs emerge. You won’t lose your job to AI; you’ll lose it to someone who uses AI better than you. Upskill. Prompt better. Stay relevant.
Nidhin put it more honestly: “That’s basically a veil. Enterprises are so diverse in their systems that making sense of a business is still slightly difficult for an independent AI agent. At this stage. But when that changes — and it will — the orchestration itself becomes AI.”
And when that happens, what’s left? The jobs that require you to be accountable for something. To sit in the room when it goes wrong. To make a call no one can offload to a model.
Those jobs are fewer. They don’t scale.
Nidhin also raised something that didn’t get enough airtime — what mass displacement does to economies like India. When cities can’t guarantee clean water and power but are being asked to host data centers for foreign AI companies? When the youth unemployment that sparked the Bangladesh uprising is echoed across the subcontinent? When a government tells you the economy is great and simultaneously raises fuel prices?
These aren’t separate stories. They’re the same story.
One More Thing
The week we recorded this episode, Anthropic — the company I use every day, and trust less than I probably should — publicly called for a global pause on AI development. Their co-founder wrote that without coordinated slowdowns, the race would be driven by whoever is least cautious.
Their own AI Safety Head had already resigned in February, saying publicly: “the world is in peril.”
These are not outsiders raising the alarm. These are people who built the thing.
I said it in the episode and I’ll say it again here: “I don’t trust Anthropic. I don’t trust any of these AI companies. But it’s interesting that the one who keeps saying ‘slow down’ is the one who is also going public and taking the money.”
Today, SpaceX rang the opening bell on Nasdaq.
Tomorrow, it’ll be someone else’s turn.
And we’ll keep watching — and talking — until they come and get us.
Episode 11 of The Sunday Draft — “Are We Enabling AI to Replace Us?” — is out now on YouTube, Spotify, Apple Podcasts, and here on Substack.
Subscribe for new episodes every Sunday.