AI Researcher Andrew Tulloch Leaves Meta, Joins Anthropic to Optimize Claude Code

Andrew Tulloch, a top AI researcher who led Muse development at Meta's TBD Lab, has left Meta just two days after the launch of Muse, ending his second stint at the company after about 11 months. According to sources, Tulloch will join Anthropic's reasoning and performance team next week, where he will work on optimizing Claude Code's training pipeline and improving model response efficiency.

Tulloch was one of the most expensive hires in Meta's 2025 talent war. In August 2025, Meta CEO Mark Zuckerberg and Scale AI founder Alexandr Wang courted him with a compensation package reportedly worth around $1.5 billion. Tulloch initially declined, but two months later he left Thinking Machines Lab, a company he co-founded with former OpenAI CTO Mira Murati, and rejoined Meta to work on superintelligence research at TBD Lab.

Before that, Tulloch had spent more than 11 years at Meta, contributing to projects including PyTorch, ResNet-50 training optimization, and FBGEMM. He later joined OpenAI in 2023, working on GPT-4o, GPT-4.5, and o3.

His departure comes two days after Meta launched Muse on September 8, 2026. Muse is a personal AI agent built on the Muse Spark model, which Meta describes as its strongest model for real-world agent tasks. Earlier, another researcher recruited from OpenAI, Jiahui Yu, also left Meta after the release of Muse Spark 1.2.

Anthropic has not disclosed further details about Tulloch's role, but confirmed he will focus on training and performance improvements for Claude Code.