The global war for talent in AI reaches a new threshold with Meta, which has just attracted one of the most coveted engineers in Silicon Valley, in the person ofAndrew Tullochco-founder of the startup Thinking Machines Labfounded with Mira Murati (ex-CTO of OpenAI). The latter therefore leaves the company to join Meta Platforms, and according to the Wall Street Journalhis total remuneration could reach 1.3 billion euros over six yearsincluding salaries, bonuses and stock options.
This hiring illustrates the offensive strategy of Mark Zuckerbergdetermined to make up for Meta’s delay in the face of OpenAI, Anthropic And Google DeepMind. According to several sources, the CEO of Meta had first tried to buy Thinking Machines Lab, before approaching several of its engineers individually. Tulloch was among the dozen profiles personally solicited by Zuckerberg.
The operation comes as the global AI talent market reaches unprecedented valuation levels. The CEO of OpenAI, Sam Altmanconfirmed earlier this year that Meta had proposed up to 85 million euros bonuses to certain senior researchers to poach them. The company, long seen as a scientific follower, is now devoting a growing part of its investments to strengthening its teams, and no longer just to computing power.
In June 2025, Meta had already purchased 50% of Scale AI for €12.1 billionmainly to secure the services of its founder Alexander Wangnow at the head of Meta Superintelligence Labs. This new division focuses the group’s efforts to design artificial general intelligence (AGI) architectures. At the same time, Meta managed to recruit eleven high-level engineers from OpenAI, Anthropic and Google.
For Thinking Machines LabTulloch’s departure represents a symbolic setback. The company, still young, attracted attention by bringing together a team from OpenAI and DeepMind around hybrid models combining neural architecture and formal logic.
Andrew Tulloch: from Australian prodigy to global AI elite
Andrew Tulloch has established himself as one of the most brilliant engineers of his generation. Graduate of theUniversity of Sydney with the highest distinctions (University Medal, Joye Prize), he specialized in applied mathematics and intensity models for credit risk. In 2007, he already represented Australia at the International Chemistry Olympiad, winning a world silver medal.
After a stint at Goldman Sachs in Sydney, where he applied machine learning to market microstructure, he joined Facebook in 2012. He developed there with Ralf Herbrich one of the group’s first distributed learning systems, capable of analyzing billions of daily ad impressions and predicting click-through rates in real time.
In 2014, Tulloch continued his studies at Cambridge Universitywithin Trinity Collegewhere he gets a Master of Mathematical Statistics (Part III of the Mathematical Tripos) with distinction and the college award for academic excellence.
Between 2023 and 2024, it joins OpenAI as Member of Technical Staffparticipating in the training of GPT-4o And GPT-4.5and contributing to the design of the reasoning modules in the series o-series. In January 2025, he co-founded Thinking Machines Lab with Mira Muratibefore being recruited ten months later by Meta Platforms.
Thinking Machines Lab: the long shadow of a future rival
Founded in early 2025 in San Francisco, Thinking Machines Lab develops general artificial intelligence architectures combining deep learning and symbolic reasoning. The company had raised funds from American and Asian private investors, without public communication on its valuation. Its ambition: to go beyond simple imitation of language to build models capable of self-improvement.
Tulloch’s departure, however, weakens the young company, still in the R&D phase, at a time when investors are scrutinizing the viability of independent laboratories in the face of accelerated consolidation in the sector.
