Hollywood writers are quietly becoming one of the largest pools of contract labor training frontier AI models. In a first-person account published this week, screenwriter and showrunner Ruth Fowler describes completing 20 AI-training contracts over eight months across five platforms — Mercor, Outlier, Task-ify, Turing, Handshake and Micro1 — at rates ranging from $52 an hour for generalist data annotation to $150 an hour for expert creative writing on OpenAI projects.
The pipeline runs from the Writers Guild of America Facebook groups straight into the contractor portals. Fowler found her first gig through a post in an unofficial WGA group, where a peer wrote that Mercor was "paying 150 an hour for writers. It's easy money." The reality, she reports, was 10 job applications, roughly 20 unpaid hours of capability tests, and an interview with an AI recruiter agent before she was hired at $52 an hour six weeks later.
The work itself spans the full surface area of modern model development. Fowler grades chatbot responses against a rubric, annotates videos frame by frame, generates anime sex scenes and decapitation imagery for red-team safety testing, and probes large language models to extract bomb recipes and synthetic invites to a January 6 reprise at the White House. Workers identify themselves on the platforms only by handles like ri611 or h924092b12ee797f.
Key facts
- 01Ruth Fowler details eight months and 20 AI-training contracts across Mercor, Outlier, Task-ify, Turing, Handshake and Micro1.
- 02Generalist annotation paid $52 an hour; expert creative writing for OpenAI paid $150 an hour.
- 03Onboarding required 10 job applications and 20 unpaid hours of testing before her first paid task.
- 04A planned 20-hour-a-week, two-month project was killed without warning after she logged 10 hours a week for two weeks.
- 05A second 'expert' project paid just $180 in the four weeks before Christmas after repeated launch delays.
The economics of the contractor economy are the story. Fowler's first project was scoped at 20 hours a week for two months, but she logged just 10 hours a week for two weeks before the project was unplugged with no notice. Her project manager, a 22-year-old recent graduate, Slacked: "Sorry guys. I had no idea this was coming." The Slack channels, Airtables and onboarding documents were dismantled within hours.
“In eight months on five different platforms, Fowler completed 20 contracts, with rates ranging from $52 an hour for generalist annotation to $150 an hour for expert creative writing on OpenAI projects.”— Jaeden Schafer
Four weeks later she was offered an "expert" role at $70 an hour. The project, code-named in a dead language — Fowler observes that expert projects get classical names while annotation grunt work gets named after woodland creatures — was supposed to start within a week. It launched four weeks later at 9 pm on a Monday, plagued by Airtable registration glitches that locked workers out. By the time access was restored 24 hours later, the finite pool of tasks had been claimed. Fowler earned $180 in the four weeks before December 1.
The platforms market this volatility as flexibility. Mercor's own pitch, quoted by Fowler, reads: "Contractors on Mercor's platform choose when and how much to work. How they participate on the platform is up to them." A team leader reframed it more bluntly in a midnight Slack message, telling Fowler she should not rely on the work or expect anything from it: "These are not jobs, these are 'tasks,' and we are 'taskers.'"
The labor supply is a direct consequence of the 2023 Writers Guild strike, which lasted nearly five months and ended with contract language meant to prevent studios from replacing writers with AI. Fowler argues the entertainment industry never regained its momentum after the strike. In early 2025, after a producer defaulted on a six-figure check owed for a TV show she created, she started looking for parallel income — and landed in the same labor market the strike was meant to constrain.
A mid-level TV writer Fowler calls Jonathan is paid $150 an hour to evaluate scripts for OpenAI as an Expert Creative Writer. He describes the dynamic as "a bit Hunger Games" — workers shift their sleep schedules to match team leaders' hours and compete for assignments from the TLs who can grant or withhold work. The credentialing skews high: experts typically hold master's degrees and have significant domain experience in fields like neurology, linguistics, real estate or journalism.
The platforms named in the piece have not publicly addressed individual contractor experiences, and the descriptions reflect one writer's account across her own contracts. Mercor's standard contractor terms emphasize self-directed scheduling, and the broader RLHF-and-evaluation labor model is widely used across frontier labs to refine model outputs. The work Fowler describes — preference rating, red-teaming, multimodal annotation — is the same labor pipeline underpinning capability gains at every major lab.
The deeper signal here is structural. Frontier model training has created a large, white-collar gig economy that sits adjacent to — and increasingly inside — industries that AI is disrupting. Screenwriters, journalists, neurologists and linguists are being paid by the hour to teach models the judgment that those same models will eventually be asked to perform autonomously. The pay is real, the projects are real, and the work is foundational to model quality. It is also unstable in ways that traditional contracting is not: projects materialize and vanish in days, payment rates vary by an order of magnitude across tiers, and the labor relationship is mediated by Airtables and AI recruiters rather than HR departments. For the AI industry, that flexibility is the feature. For the workers funding their rent with it, it is the cost.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
