Particle.news

Download on the App Store

Meta’s Superintelligence Drive Stumbles Early With Exits, Scale AI Friction, and a Key Title Upgrade

Pressure is building to show faster progress, with reports pointing to a year-end target for an improved Llama release.

Overview

  • TBD Labs is sourcing training data from rivals Mercor and Surge despite Meta’s $14.3 billion investment in Scale AI, as several researchers reportedly view Scale’s data as lower quality.
  • Ruben Mayer, a former Scale AI executive brought in to help run Meta Superintelligence Labs, left after roughly two months, while Scale has recently lost OpenAI and Google as customers and cut 200 labeling jobs.
  • Meta confirms only one active TBD Labs staffer, Ethan Knight, has departed; Avi Verma never started, and Rishabh Agarwal announced his exit from MSL, as veteran leaders including Chaya Nayak and Loredana Crisan also leave.
  • Shengjia Zhao, a ChatGPT co-creator, was named Meta’s chief AI scientist after he considered returning to OpenAI, with Meta saying his title formalizes a role he already held as scientific lead.
  • Meta instituted a temporary hiring pause and reorganized MSL into focused research, product, infrastructure, and FAIR-aligned teams, following Llama 4’s lukewarm reception and reported plans to deliver a new release by year-end.