Mercor will acquire Deeptune, the New York-based creator of high-fidelity reinforcement learning environments for AI agents. Deeptune’s team, led by founder & CEO Tim Lupo, will join Mercor in NYC. The deal follows Deeptune’s $43M Series A led by Andreessen Horowitz.



Mercor, a San Francisco-based company that connects AI companies with experts to train AI models, announced on July 9, 2026 that it will acquire Deeptune, a company focused on building environments for reinforcement learning.

The acquisition marks a strategic move by Mercor to expand its role in the AI training infrastructure market, particularly as reinforcement learning becomes increasingly important for developing AI agents capable of performing complex digital work. Deeptune creates high-fidelity simulations of the digital world, providing sandbox environments where AI agents can practice useful tasks, receive measurable feedback, and improve through reinforcement learning.

Brendan Foody, co-founder and CEO of Mercor, said the industry’s constraint is shifting from model capability alone to the availability of realistic environments where models can practice and be evaluated.

“Reinforcement learning has reached the point where a model can learn almost any task that can be clearly defined and scored,” Foody stated in a blog post. “The constraint has shifted to the environments themselves: the places where models practice the work and get measured on whether they did it well.”

As part of the transaction, Deeptune’s team of engineers and operators, including founder and CEO Tim Lupo, will join Mercor’s team in New York City. Financial terms of the acquisition were not disclosed.

The deal brings together Mercor’s expert network and Deeptune’s software platform for reinforcement learning environments. Mercor’s experts already contribute to AI training by writing evals, while Deeptune provides the technical infrastructure to convert that expertise into realistic training environments at scale.

“Mercor’s experts are already doing much of this work by writing evals,” Foody said. “Deeptune provides the missing piece: the software platform where that expertise becomes realistic training environments at scale.”

Deeptune previously announced a $43 million Series A funding round in March 2026. The round was led by Andreessen Horowitz, with participation from 776, Abstract Ventures, Inspired Capital, and angel investors including Foody; Noam Brown, a researcher at OpenAI; and Yash Patil, CEO of Applied Compute.

The acquisition comes as AI companies increasingly focus on reinforcement learning and evaluation-driven training methods to improve AI agents’ ability to perform real-world tasks. For enterprise applications, including recruiting, HR operations, customer support, finance, and workflow automation, AI agents must be trained and tested in environments that closely resemble actual business systems and processes.

By integrating Deeptune’s reinforcement learning environment platform, Mercor is positioning itself to support a more complete AI training loop: expert-defined tasks, evaluation design, realistic simulation environments, agent practice, measurable scoring, and continuous model improvement.

For the broader AI and HR technology markets, the transaction highlights an important shift in how AI capabilities may be developed and validated. As companies move from experimental AI tools toward autonomous agents embedded in business workflows, the ability to create scalable, realistic, and measurable training environments could become a critical layer of enterprise AI infrastructure.