Google and Meta race to build personal AI agents as Anthropic and OpenAI pull further ahead

The shelving of Google's Mariner browser agent isn't just a product cancellation. It's an admission that browser-layer automation lost the architectural argument to deeper OS and app-level integration, which is a meaningful strategic concession from a company that built Chrome.

Meta's situation here is worth reading alongside its acquisition of Assured Robot Intelligence from early May. That deal positioned Meta as a platform layer in robotics, and the personal agent push follows the same logic: own the integration surface rather than compete on model quality alone. For Google, the pressure is compounded by the Pentagon AI deals covered around May 1st, where Google secured classified contracts that OpenAI also landed. Winning government infrastructure while losing the consumer agentic narrative to Anthropic and OpenAI creates a split identity problem that is hard to manage from a single product org. Anthropic, notably excluded from those Pentagon deals after friction over usage terms, has instead been building domain-specific deployment channels like Claude Security, which suggests a deliberate choice to compete on controlled vertical depth rather than broad platform reach.

Watch whether Meta ships a consumer-facing personal agent product before Google replaces Mariner with a named successor. If Meta moves first, it confirms the robotics acquisition and agent strategy are part of a coordinated platform play rather than parallel bets.