Google closes $1.5 billion talent and licensing deal with coding startup Mechanize

Google has quietly finalized a high-profile transaction valued at over $1.5 billion to absorb the core research team and license the technology of Mechanize Inc., a San Francisco-based startup specializing in high-fidelity reinforcement learning environments for coding models. The transaction is structured as a non-exclusive technology licensing agreement paired with the mass hiring of Mechanize's technical staff. This playbook has become the tech industry's preferred method of executing corporate roll-ups while bypassing the intense antitrust reviews associated with traditional full-scale acquisitions.

As part of the agreement, Tamay Besiroglu, the co-founder and former CEO of Mechanize, has officially joined Google’s AI division, DeepMind, as a Research Scientist. Besiroglu, a former MIT FutureTech researcher and associate director of the research group Epoch AI, is bringing over a dozen of his core engineering and technical staff to Google with him. The incoming team is reportedly being integrated into Google DeepMind’s 'midtraining' division—the critical phase of training where models are refined using reinforcement learning from human and machine feedback to handle long-horizon agentic workflows.

While the bulk of the technical team is transitioning to Google DeepMind, Mechanize Inc. will continue to operate as an independent entity under the leadership of former Chief of Staff Guive Assadi, who has assumed the role of Chief Executive Officer. Google's licensing fee will be used to resolve the startup's existing liabilities and compensate its initial seed investors, who backed Mechanize in April 2026 with a $9.1 million funding round. The rapid transaction highlights the fierce corporate battle to control the systems that evaluate and grade autonomous coding models.