Anthropic acquires Vercept to boost Claude’s ability to operate software like humans

On February 25, Anthropic announced its acquisition of Vercept, aiming to enhance the capabilities of its Claude models in executing complex tasks within software applications. This strategic move highlights the increasing competition in the AI landscape to develop systems that can operate autonomously within digital workflows. The acquisition reflects a broader trend among AI developers who are transitioning from basic chatbot functionalities towards creating 'computer-using' agents. These agents are designed to navigate live applications, manage multi-step workflows, and coordinate tasks much like a human employee would. While Anthropic has not disclosed the financial details of the deal, the implications are significant. "Users are turning to Claude for more intricate tasks, including writing and executing code across large repositories, synthesizing research from various sources, and managing complex workflows that involve multiple tools and teams," stated Anthropic in a recent announcement. The ability to use software interfaces, rather than relying solely on code execution or textual responses, empowers Claude to perform tasks in live applications as a human would at a keyboard. This capability means Claude can undertake multi-step tasks in real-time applications and tackle challenges that are not solvable through code alone. Vercept specializes in enabling AI systems to perceive and interact with graphical user interfaces, addressing a significant technical barrier in creating autonomous digital agents. Anthropic emphasized that solving complex perception and interaction issues is essential for making AI genuinely useful for intricate tasks. The founders of Vercept, Kiana Ehsani, Luca Weihs, and Ross Girshick, along with their team, will join Anthropic, while Vercept plans to phase out its external products in the next few weeks. Anthropic noted, "The Vercept team has dedicated years to understanding how AI can see and function within the same software environments that people use daily," adding that their expertise directly contributes to addressing some of the most challenging problems the company faces. This acquisition follows the recent launch of Claude Sonnet 4.6, which has reportedly enhanced the model's proficiency in executing tasks across operating systems and web environments. Anthropic revealed that its Sonnet models have improved dramatically on OSWorld, an AI performance benchmark, rising from under 15% in late 2024 to an impressive 72.5% today. This advancement is bringing them closer to achieving human-like proficiency in tasks such as navigating spreadsheets and completing web forms across multiple browser tabs. Vercept represents the latest addition to Anthropic's series of acquisitions focused on accelerating its technological development. The company is actively seeking teams whose technical aspirations align with its own, particularly in the realm of safety and usability.