I use aliases like ri611 or a long hashed username on the platforms where I work, trading anonymity for short bursts of income. In another life I was a Hollywood writer and showrunner, creating prime-time drama for networks and streamers; the 2026 strike and an unpaid six-figure check in early 2026 nudged me toward alternative income. What I discovered was a hidden labor system that asks creative professionals to teach machines while subjecting them to the rhythms of the gig economy. The tasks can be creative and grotesque, technical and trivial, and they are usually frustratingly intermittent.
My contracting roles have involved rating chatbot tone, annotating furniture photos, removing strangers from group portraits, time-stamping tiny events in videos, and testing safety limits by simulating provocative prompts. I worked with companies named Mercor, Outlier, Task-ify, Turing, Handshake, and Micro1. The job title is often AI trainer or annotator, and the work is delivered through a stack of apps: Slack, Airtable, Zoom rooms, and opaque payment portals. The experience is less a steady career than a sequence of one-off assignments that can vanish without warning.
The mechanics of the work
Most assignments begin with multiple stages of vetting: assessments, unpaid tests, and an interview conducted by an automated recruiter or a flickering video window. I remember finally getting an offer in September 2026 after many applications, dozens of unpaid hours, and a qualifying chat where I judged AI prose. Entry as a generalist might pay around $52 an hour; occasionally you are elevated to an expert slot that commands higher rates—I was once offered $70 an hour, and at the industry peak some projects advertised $150 an hour for specialists. But advertised rates are unstable and projects are often called sprints, short bursts of labor that can pop into existence and evaporate within days.
Promises, platforms, and the churn
What the platforms sell to workers is freedom: set your hours, pick your tasks, work when you like. In practice, the promise is undermined by constant onboarding, mandatory Slack check-ins, and surprise project shutdowns. I experienced project starts and abrupt cancellations firsthand: a two-month engagement scheduled for 20 hours a week was unplugged without notice after a couple of slow weeks; another project called Dead Language promised holiday earnings but launched so late the finite batch of tasks disappeared within a day. Between February and April 2026 I was hired and off-boarded across seven projects on four platforms—each termination abrupt and usually unexplained.
Onboarding and the hidden gatekeepers
Onboarding often feels like a maze of quizzes and sign-ups. You join a Slack, register in Airtable, and wait in long Zoom rooms where faceless helpers patrol support squares. If you miss a single form or repeat a signup, you can be locked out; I once spent days chasing access only to be removed after failing to click the right quiz link. Many contractors are told to think of this work as a bonus or a second job, but for some it is the main source of income. The rules enforced by team leaders—usually young graduates—can determine your continued eligibility, while non-disclosure agreements deter public discussion.
Metrics, badges, and competitive pressure
Performance is quantified aggressively: reviewers grade annotations on a five-point scale, track your average handling time, and reward top performers with visible badges that purportedly help secure future work. These systems create frenzy when a new batch is released: annotators race to capture scarce tasks, sometimes working through the night and sacrificing family time. Scoring is granular to the point of absurdity—tiny word changes or phrasing flags can tip a result from a 5 to a 1. That volatility, combined with shrinking wages (I saw generalist rates slip to the teens per hour), fuels burnout and anxiety.
Community response and legal friction
Workers have turned to Reddit, Discord, and forums to compare notes, vent, and organize. In November 2026 thousands reportedly lost jobs or faced pay cuts when projects were replaced with lower-priced equivalents; some who relied on that income were left scrambling. The debate has spilled into courts: lawsuits allege misclassification of contractors, arguing that the cadence of mandatory tasks, mandatory check-ins, and de facto scheduling resemble employment. Community threads are often raw—people share horror stories of being rehired at a lower rate, or of being fired for falling ill. The solidarity is real, but so is the fear of retaliation under NDAs.
What this means for workers and ai
The industry promises to make machines more human by forcing humans to mimic perfect behavior under pressure. As platforms optimize for speed, precision, and lower cost, they shape a workforce that must be faster and cheaper each cycle. For creative professionals like me, the transition from scripted narrative to annotation revealed a harsher narrative arc: irregular pay, opaque rules, and a relentless emphasis on metrics. Until labor practices catch up with this new form of work, many skilled people will keep rotating between creative careers and the precarious underside of data annotation, trying to pay rent and keep their dignity intact.