5000 Laid Off Overnight! Mercor Rehires at Lower Pay — Unbelievably Shameless

Honghao Wang

29 Nov 2025 — 3 min read

Xinzhiyuan Report

---

Introduction

They are the hidden parents of AI — teaching machines to recognize emotions and understand language — yet some were discarded overnight. One case: Mercor shut down its Meta project, leaving 5,000 data labeling workers jobless, then invited them back at lower pay.

In the wave of the AI revolution, machines upgrade but human value declines.

---

The Relentless AI Boom

Tech giants are racing to invest heavily in AI.
Analysts claim AI is responsible for 90%+ of U.S. GDP growth.
Media paints a vision of a "new era of human prosperity".

Yet, away from the celebrations, quiet mass exits happen.

Last week, one single email from Mercor ended 5,000 jobs overnight — revealing part of the hidden cost of the AI surge.

---

Overnight Layoffs & “Pay-Cut Restart”

What Happened

Mercor, a data labeling provider for OpenAI, Anthropic, and Meta, suddenly ended its contract for the Musen Project.

Key points:

Timing: Mid-November termination.
Scale: ~5,000 workers affected.
Before: Employees were assured the contract would extend.
After: Email arrived, work was gone.

Days later, Mercor announced the Nova Project:

Same work as Musen.
Lower pay — $21/hour reduced to $16/hour.
Marketed as "more stable" and "more flexible."

> Irony: Mercor had just closed a new funding round with a valuation near $10B.

Industry Pattern

This cycle — contract ends, mass layoffs, immediate rehire at lower rates — is common.

Companies involved: xAI, Scale AI, Appen.

The difference: Mercor was unusually open about it.

---

Life on the AI Assembly Line

Workers’ voices:

> “They said the client was happy, the project would renew. Then suddenly it was terminated.”

> “At 3 a.m., I got the email. Next morning, my task page was gone, replaced with ‘Thank you for your contribution.’”

Later:

Invitation to Nova project.
Pay drop from $21 → $16/hour.
Corporate language about "task stability" and "efficiency."

Why they accepted:

Need for stable income.
Obligations: children, student loans, lack of alternatives.

Conclusion: Same pipeline, less pay — humans downgraded even as AI systems improve.

---

The Bigger Picture: Shrinking Human Value

AI companies thrive; workers are sidelined.

Economic data:

Tech contributed 92% of US GDP growth in H1 2025.
Highest tech layoff numbers since 2003.
Amazon cut 14,000 corporate staff.
Google and Meta continue “team optimization.”

CEO Narratives vs Reality

OpenAI’s Sam Altman:

> "We are stepping into an era of human prosperity beyond imagination."

But on the ground, data annotators — the builders of AI models — aren’t part of this “prosperity.”

---

"Flexible Employment" — The Euphemism

Workers labeled “freelancers” or “independent collaborators.”
Reality: No insurance, no paid leave, no job security.
Disconnection happens via email or expired log-in link.

Paradox: To seek stability, annotators must rely on unstable work.

AI firms:

Replace employment with collaboration.
Mask layoffs with “flexibility”.
Manage via algorithms.

Workers continue to make AI smarter — even as their own names vanish from the system’s roster.

---

Possible Paths Forward for Human Contribution

Platforms like AiToEarn官网 demonstrate alternative models:

Open-source global AI content monetization.
Integration of AI generation, cross-platform publishing, analytics, and model ranking (AI模型排名).
Publish simultaneously across Douyin, Kwai, WeChat, Bilibili, Rednote (Xiaohongshu), Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, X (Twitter).