Not sure if you have noticed, but you are in a cult.
Not a robed one. A professional one. It has a name for its members (contributor, fellow, expert), a private language (AHT, R2I, rubric, gate), a shared enemy (the models that get everything wrong), and a promise that the work you are doing matters more than the pay rate suggests. It has initiation rituals that cost you unpaid hours. It has channels where you can see everyone else’s devotion but nobody’s invoice. And it has the one feature every group like this shares: leaving feels like losing something you cannot get back.
I want to name it plainly, because I am part of it too. I have worked more than ten Mercor projects on both the contributor and the reviewer side. I read the Slack channels. I know the feeling of a queue that goes quiet for three weeks and the relief when it comes back. That relief is the mechanism. It is the thing that keeps very smart people accepting whatever arrives.
This post is about how the expert-data economy works on you, what a sane remote working life looks like inside it, and what I do for paid subscribers who want somebody on their side of the table. Most of all it is about being free inside this jungle, which is not the same as leaving it.
Why the AI expert economy feels like a cult
The playbook for building a movement is well understood. You need a compelling story about the future, an in-group with its own vocabulary, a sense that outsiders do not get it, and a founder who can articulate why the mission justifies the sacrifice. Every one of those pieces is present in AI data work.
The story. The pitch is that the next decade of knowledge work is about encoding your expertise into models. Mercor’s CEO has said the network passed 4 million vetted experts and pays out more than $2 million a day. Handshake AI reports $100 million paid to more than 100,000 fellows. Frontier labs are spending over $1 billion each on training environments. None of that is fake. The money is real and the mission is real. That is exactly what makes the rest of it work.
The initiation. Onboarding and assessments are unpaid on most platforms. You give hours before you see a rate. Many get onboarded, few get retained, and the gates that decide who stays return no score and offer no appeal. You learn to treat acceptance as a gift instead of a contract.
The vocabulary. Once you know what AHT means, you are inside. Maximum handling time caps the paid hours per task no matter how long the task takes. A nominal $60 to $100 an hour becomes an effective $8 to $25 once the cap bites. On Handshake’s Project HH in May 2026, workers were paid 20 to 50 percent of the hours they logged. I covered the mechanics in The AHT Game and the Handshake failures in the Handshake issue. Inside the group, these become normal facts of life. From outside, they look like what they are.
The silence. Nobody tells you why the queue stopped. Nobody tells you why the radiologist got a music-tagging task. Silent offboardings run at a roughly 7.5 percent rate on one platform, with no appeal. You fill the silence with your own explanations, and most of the explanations you reach for are about you.
I want to be careful here. I do not think this is designed by villains. It is a sieve. The platforms are built to select for people who tolerate uncertainty, unpaid ramp time, and rate compression without complaint. The people who stay are the people the system fits. That is a selection effect, not a plot. But the result for you is the same either way: you adapt to the sieve instead of the sieve adapting to you.
What the paid tier actually is
It is this newsletter’s verification work, done for one profile instead of a public list. Yours.
You tell me where you are. I go find what fits, and I tell you what I would do if I were you, in addition to the paid posts.
1. The profile prompts. The Mercor one that produced the offer above, plus versions for micro1, Handshake AI, and the smaller platforms. Updated when the platforms change, which they do.
2. Matched openings, sent to you weekly. AI training and expert-data roles across the big platforms and the small ones, filtered against your background, your region, and your license. Including postings that never surface on aggregators because they close before anyone indexes them.
3. Advices for remote work operations. Running a distributed week, handling tasks across a nine hour gap, managing Slack so it does not manage you, tracking your own hours independently so you can dispute a shortfall in the same week it happens.
4. The offboarding-risk playbook. Where the handling time corridors sit on projects I have run, what the assessment gates actually test, and how to exit a project without it costing you the next one. Built from more than ten projects on both the contributor and the reviewer side.
I cover Mercor, Handshake AI, and micro1 in the same place, which matters more than it sounds like. They are not substitutes. Mercor pays higher at the top and goes quiet for weeks at a time. Handshake is steadier and lower, and has its own payment problems that I have documented and will keep documenting. Holding both queues open is how a bad month on one stops being a zero month.
How to start, in one reply
Upgrade here, then reply to the welcome email with three lines:
Which platform or platforms you are on.
Which stage you are at: not applied, applied and waiting, onboarded and idle, or actively on a project.
Your background, your region, and any license or specialty.
That is enough for me to start matching. You get the profile prompt back the same week, and your first matched roles in the next customized email.
Below the wall: the habit system that makes remote work run without a boss, and the monthly focus exercise I use when the platform silence starts writing stories in my head at 3am.
