Physix Frontier · News Briefing Card (enbrief · Sep 24, 2026)

AI Data Labeling Shifts to High-Paid Experts, Firms Face Double Harvest

KEY FACTS

Amanda Brown, an assistant professor in Texas, took gig work on a data labeling platform at $60 per hour and quit after a few months.
Mercor pays over $4 million per day to 30,000 outsourced workers, recruiting experts such as physics PhDs and doctors.
After pivoting to data training, Handshake's annualized revenue surpassed $1 billion.
Microsoft CEO Nadella proposed the reverse information paradox, saying users leak proprietary knowledge to use AI.
Microsoft cut about 4,800 jobs, while Cloudflare laid off over 20% despite revenue growth exceeding 30%.

KEY DATA

60美元Data labeling hourly wage

400万美元Mercor daily wage expenditure

3万人Mercor outsourced workers

10亿美元Handshake annualized revenue

PHYSIX OBSERVATION

AI companies are buying out expert knowledge with high salaries, distilling human experience into commoditized models. Enterprises feed their core know-how for efficiency, but may end up with a mediocre general brain. Both individuals and companies should ask: why should my scarce knowledge be cheaply traded to giants? Building private evaluation and trust boundaries is the real moat.

Source: enbrief original report ↗