Abi Olvera
While working on a research paper, I decided to test one of the leading AI assistants and asked Anthropic’s Claude to analyze hundreds of emails and build a spreadsheet of recent Nobel Prize-winners. Claude delivered, referencing websites, organizing data, and catching some, though not all, of its own errors.
Then I asked the app to schedule a simple dinner gathering between several colleagues.
It botched the start time twice.
This is generative AI in 2025—systems that can synthesize information across websites, email, and spreadsheets with sophistication but trip up with something as simple as calendar event start times. Nonetheless, it represents a huge advance: The ability to connect AI systems directly to one’s online life wasn’t possible until several months ago. To get AI assistants to interact with other web services, like reading an email inbox, most users needed a third-party program or browser specific to their use-case. Now, Claude and other AI models can connect directly with email providers to read and respond to messages, set up calendar events, and create and edit spreadsheets. Claude can even connect with popular task and project management systems like Asana and Notion, as well as payment processing tools like Stripe. This allows AI “chatbots” to become “AI agents,” or models that interact with the real world and can carry out tasks.
What enabled these integrations was the development of a framework called model context protocol (MCP)—a small technical standard that has quietly become the invisible infrastructure connecting AI to the rest of our digital world. But this protocol wasn’t built for what it’s doing now. It was designed for a very narrow and basic purpose: letting Claude’s desktop app connect to local files and simple tools. Its widespread adoption introduces some opportunities, like accelerating cybersecurity agents, but also some privacy and security challenges. The tech industry and open-source community are driving efforts to solve these issues, though the pace of adoption makes keeping up difficult.
No comments:
Post a Comment