A Nation feature published April 14 makes a blunt case: the same venture capitalists who pushed to gut US science funding are also backing the gig platforms hiring displaced PhDs to train AI models at $30 to $90 an hour. Reporter Hirsh Chitkara interviewed researchers who turned to services like Mercor after their federal funding dried up.
The VC loop
Peter Thiel and Marc Andreessen spent years arguing that academic science had grown bloated. In leaked texts obtained by The Washington Post last year, Andreessen called for the National Science Foundation to face "the bureaucratic death penalty." Thiel, in a Hoover interview, claimed the average PhD today is "99% less productive" than a century ago, an argument that glides past the possibility that running a particle collider might take more paperwork than a Bunsen burner.
The cuts arrived on schedule. The administration's proposed FY2026 budget, tracked by the AAAS dashboard, called for slashing NIH by 40%, NSF by 57%, and NASA by 24%. Congress clawed some of it back. Damage was already baked in: Science reported more than 10,000 federal STEM PhDs left government in the past year. University labs, per The Atlantic, shed staff or shut down entirely.
Thiel-backed Mercor just raised at a $10 billion valuation. Meta bought a 49% stake in Scale AI, also a Thiel investment, at a $29 billion valuation. Both pitch themselves as the marketplace for "PhD-level" training data. Both need the same researchers whose jobs just disappeared.
$90 an hour, with an asterisk
Ads for these platforms lean hard on Uber's old playbook. Academics hiking, reading in hammocks, voiceovers about flexibility and staying in the field. "Finding jobs in academia has always been a struggle," says one contractor in a Mercor testimonial, before explaining he joined up after his summer funding got cut. The framing writes itself: not unemployment, freedom.
Numbers tell a different story. One applied math doctoral student told Chitkara he was assigned tournament-level problems at a nominal $90 an hour, but the platform capped pay at 2.5 hours per question and paid nothing for incomplete or wrong answers. An MIT engineering PhD said she started logging all the hours she wasn't being paid for. "It ended up not being super worth it," she said, which is the most polite possible way to describe a rate that quietly collapses once you do the math.
Brendan Foody, the 23-year-old Mercor founder, described his company as the "primary bottleneck" on AI model improvement in a recent interview. That inverts who's doing the bottleneck work. Researchers write the prompts, solve the problems, and eat the unpaid hours. The platform takes the billion-dollar valuation.
What the piece doesn't settle
Chitkara's interviewees drew a direct line from funding cuts to gig work. Whether that line holds at scale is harder to prove from anecdotes. The piece doesn't offer a count of how many displaced scientists are now on these platforms, or what share of Mercor and Scale's labor pool comes from former federal or academic jobs. The platforms aren't publishing those numbers either.
The next marker is the FY2027 budget cycle, when Congress decides whether to restore more of what was cut or let the PhD-to-gig-work pipeline harden into a permanent feature of US research.
