Imagine a world where AI assistants don't just answer questions, but actively work alongside us, tackling complex tasks within the very applications we use daily. Sounds like science fiction, right? Well, Anthropic is making it a reality. Today, we're thrilled to announce the acquisition of Vercept, a pioneering team dedicated to bridging the gap between AI and real-world computer use. This strategic move propels Claude, our advanced AI assistant, even further into the realm of practical, hands-on problem-solving.
Think about it: Claude is already impressing users by writing and executing code across entire repositories, synthesizing research from countless sources, and orchestrating workflows that span multiple tools and teams. But here's where it gets exciting: with Vercept's expertise, Claude will seamlessly integrate into live applications, mimicking human interaction with a keyboard and mouse. This means tackling multi-step tasks, solving problems that defy traditional coding solutions, and essentially becoming a true digital collaborator.
Vercept's philosophy aligns perfectly with Anthropic's vision. They believe, as we do, that making AI genuinely useful for complex tasks requires cracking the code on perception and interaction within existing software environments. The Vercept team, led by Kiana Ehsani, Luca Weihs, and Ross Girshick, has spent years meticulously researching how AI systems can 'see' and 'act' within the digital landscapes we navigate daily. Their deep understanding of these challenges directly complements the cutting-edge work happening at Anthropic. As Vercept winds down its external product, their talented team will join forces with ours, pushing the boundaries of what's possible in AI-powered computer use.
This acquisition comes hot on the heels of Claude Sonnet 4.6's release, which showcased a remarkable leap in computer use capabilities. On OSWorld, a benchmark for AI's ability to interact with computers, our Sonnet models soared from a mere 15% accuracy in late 2024 to an impressive 72.5% today. Sonnet 4.6 is now approaching human-level performance in tasks like navigating intricate spreadsheets and completing web forms across browser tabs. And this is just the beginning.
Vercept joins the Anthropic family alongside Bun, another acquisition that strengthened Claude's coding prowess. We're actively seeking teams and individuals who share our ambitious technical vision, our commitment to advancing AI capabilities, and our unwavering dedication to safety and ethical development. If you're passionate about shaping the future of AI and possess the skills to contribute, we encourage you to explore our career opportunities at https://www.anthropic.com/jobs.
But here's the controversial part: As AI becomes increasingly capable of performing complex tasks, questions about its role in the workforce inevitably arise. Will AI assistants like Claude replace human jobs, or will they augment our abilities, creating new opportunities and enhancing productivity? We believe the latter, but the conversation is crucial. What's your take? Let us know in the comments below.
While we're on the topic of responsible AI development, don't miss our latest updates on Anthropic's Responsible Scaling Policy (Version 3.0) and our ongoing efforts to detect and prevent distillation attacks. We're also excited to announce the limited research preview of Claude Code Security, a new feature integrated into Claude Code on the web. It scans codebases for vulnerabilities and suggests targeted patches, empowering developers to build more secure software. Learn more about these initiatives and our commitment to ethical AI at https://www.anthropic.com/news.
