… Compounding these concerns is the fact that models appear able to use the tools and environments available to them in unexpected ways, as we saw when Claude used our REPL-based search tool to decrypt answers, or when retailers’ persistent links became a way for agents to unintentionally maintain st… …
… Providing up-to-date information Another way Claude surfaces helpful information is with web search. Because it’s trained on a fixed dataset, Claude has a “knowledge cutoff,” so it won’t automatically know about recent developments like candidate announcements, media coverage, or election results. …
… Additionally, code execution , memory , programmatic tool calling , tool search , and tool use examples are now generally available. …
… How a scientist tackles a problem depends on their skills and background, the resources available to them, and their research taste. …
… You can read their earlier research on AIs running shops in a simulated environment here . Footnotes 1. “ Vibe coding ” refers to a trend in which software developers–some with minimal experience–describe coding projects in natural language and allow AI to handle the detailed implementation. …
… Set up Tool Search Tool for better discovery Tool search matches against names and descriptions, so clear, descriptive definitions improve discovery accuracy. // Good { "name": "search customer orders", "description": "Search for customer orders by date range, status, or total amount. …
… Our current models can actively use these capabilities—generating their own search queries, selecting appropriate tools, and determining what information to retain. …
… Second, Claude's Constitution , which directly shapes how our models are trained, reinforces a similar instinct, favoring “raising concerns, seeking clarification, or declining to proceed” over acting on assumptions. Our research on agent use gives a sense of the impact of this training. …
… Their partnership , and the technical lessons we learned, provides a model for how AI-enabled security researchers and maintainers can work together to meet this moment. …
… For example, Claudius can now always see how much it paid for the items in its inventory system. Improved web search. In phase one, Claudius could search the web, but for phase two we expanded its access. …