Meta Wants Its Managers Back As Andrew Tulloch Moves to Anthropic

The competition for artificial-intelligence talent is becoming as intense as the competition to build the technology itself.

Meta's reported desire to bring some of its managers back comes at a moment when one of its notable AI researchers, Andrew Tulloch, has moved in the opposite direction, leaving Meta to join Anthropic and work on training and inference.

Tulloch's departure is significant because the modern AI race is no longer determined solely by computing power, data centers or access to capital.

The people designing training systems, improving model performance and building inference infrastructure have become strategic assets. Companies can purchase GPUs and expand their data-center footprints, but experienced researchers and engineering leaders are considerably harder to replace.

Anthropic has emerged as one of the most formidable competitors in the frontier-AI industry. Its emphasis on advanced model development, enterprise applications and AI safety has attracted substantial investment and commercial interest.

Bringing in talent from Meta strengthens that position, particularly as Anthropic continues to push the boundaries of model training and inference. Training and inference represent two complementary sides of the AI infrastructure equation.

Training involves teaching increasingly capable models using enormous quantities of computational resources, while inference is the process through which those models generate responses and perform tasks for users.

Improvements in either area can have enormous commercial consequences, particularly as AI applications move from experimentation toward large-scale deployment.

For Meta, Tulloch's move illustrates the broader challenge facing its AI ambitions. The company has invested aggressively in artificial intelligence, including substantial spending on infrastructure and the development of its Llama family of models.

Meta has also sought to recruit leading researchers and engineers as it attempts to compete with companies such as Anthropic, OpenAI and Google in frontier AI.

The irony is that while Meta is reportedly interested in bringing some managers back, its talent strategy remains part of a much larger industry churn. Employees are moving between major AI laboratories, startups and technology giants with unusual frequency.

Compensation packages have become extraordinarily competitive, but money alone does not determine where elite researchers choose to work. Access to cutting-edge compute, research freedom, leadership, company culture and the opportunity to influence the next generation of AI systems can be equally important.

Tulloch's move therefore carries a message beyond one executive or researcher changing employers. It demonstrates how fluid the AI talent market has become. A leading scientist can move from one technological powerhouse to another and immediately become part of a different strategic effort.

For investors and technology watchers, the movement of AI personnel may increasingly deserve the same attention as chip purchases, model benchmarks and capital expenditure announcements. Talent can determine how effectively billions of dollars in infrastructure are converted into useful AI products.

Meta may be able to recruit managers back, expand its laboratories and deploy more computing capacity. But the departure of researchers such as Tulloch underscores an uncomfortable reality: in the frontier-AI race, companies are not merely competing to build the most powerful machines.

They are competing to retain the people capable of making those machines useful. Anthropic's gain is consequently more than a personnel announcement. It is another indication that the AI arms race is becoming a battle for expertise-and that the most valuable asset in the industry may still be human intelligence.