Jason Wei, Chain-of-Thought Researcher, Leaves Meta’s TBD Lab to Join Mercor as Chief Scientist

Jason Wei, co-author of the chain-of-thought prompting research and creator of SimpleQA, BrowseComp and HealthBench, has left Meta's TBD lab to join AI data company Mercor as chief scientist, leading its research on data and evaluations.

Christopher Ort
By Christopher Ort · 2 min read
Share
A cardboard box of books on an office desk by a window.
Image: AI-generated, i10X
On this page
  1. Quick Take
  2. In plain English
  3. What happened
  4. Background
  5. Why it matters
  6. What's still open
  7. Sources
  8. i10x Perspective

Jason Wei, a co-author of the influential chain-of-thought prompting research, has left Meta’s TBD lab to join AI data company Mercor as chief scientist. Mercor CEO Brendan Foody and Wei announced the move on X on 17 November 2026.

i10X Quick Take

Quick Take

  1. Who: Jason Wei, previously at Google Brain, OpenAI and, for about a year, Meta’s TBD lab.
  2. New role: chief scientist at Mercor, building its research team and shaping its work on data and evaluations.
  3. Known for: chain-of-thought prompting, plus the SimpleQA, BrowseComp and HealthBench benchmarks.
  4. Why Mercor: Wei wrote that “Data will shape the next frontier as more companies train AI models than ever before.”

In plain English

Mercor pays experts such as doctors, lawyers and engineers to help train and test AI models. It has hired one of the best-known researchers in the field, whose work helped AI models reason step by step. The move suggests that good training data and good tests are now seen as just as important as model design.

What happened

Foody wrote: “We are excited to bring on @_jasonwei as Chief Scientist at @mercor.” He described Wei as known “for pioneering work on reasoning and agents at Google Brain, OpenAI, and Meta” and said Wei “has been one of our favorite customers and thought partners.” As chief scientist, Foody wrote, Wei “will help build Mercor’s research team and shape our work on data and evaluations.”

Background

Wei co-authored the 2022 paper on chain-of-thought prompting, in which a model writes out intermediate reasoning steps before answering. Foody’s post calls SimpleQA “a widely-used pretraining benchmark,” BrowseComp “the top deep research benchmark” and HealthBench “the first public benchmark to popularize rubrics.”

Why it matters

For AI labs, it shows senior researchers moving toward the data and evaluation companies that labs depend on.

For Mercor, a chief scientist with benchmark experience strengthens its pitch to sell evaluations to companies training their own models.

For Meta, it loses a prominent researcher from its TBD lab.

What’s still open

  • What research Mercor will publish under Wei.
  • Whether more lab researchers follow him to data companies.

Sources

  • Brendan Foody, post on X, 17 November 2026
  • Jason Wei, post on X, 17 November 2026
  • TokenPost, “Mercor Names Jason Wei Chief Scientist to Expand AI Research”

i10x Perspective

Benchmarks and training data decide which models look good and which get better. Hiring the author of BrowseComp and HealthBench puts Mercor closer to setting those standards. More on our Meta company page and in the Business section.

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

Timeline

What happened before

  1. Release

    Muse Spark Open

    1M context · open weights

    research.meta.ai (source, opens in a new tab)
  2. Release

    Muse Wonder

    1.05M context · $4 / $20 per 1M tokens

    Meta's most capable models to date for agentic work and coding; Muse Wonder via Meta Model API and Muse Code, Muse Wonder Full for Muse Code subscribers.

    about.fb.com (source, opens in a new tab)
  3. Release

    Muse Spark 1.1

    ai.meta.com (source, opens in a new tab)
  4. Release

    Muse Spark

    First model from Meta Superintelligence Labs; available in meta.ai and the Meta AI app.

    ai.meta.com (source, opens in a new tab)
  5. Partnership

    Nebius signs up to $27B AI infrastructure deal with Meta

    With Nebius · Compute · $27B

    nebius.com (source, opens in a new tab)
Full Meta timeline ›

Editor's picks

  1. Remanence Raises $30 Million Seed to Turn Companies’ Agent Data Into AI Models They Own
  2. Astra for Law: OpenAI Opens Its Legal AI to All ChatGPT Enterprise Customers
  3. Proton Lumo Desktop Brings a Privacy-First AI Assistant to Windows, Mac and Linux
  4. GitHub Copilot Evals Lets Developers Test Their Own Agents as Code Review Adds Adaptive Effort and Secret Detection
  5. Copilotforce: Microsoft and Salesforce Bring CRM Data and 38 Sales Skills Into Copilot